API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Sep 22nd, 2017
67
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Diff 105.58 KB | None | 0 0
raw download clone embed print report
  1. diff --git a/.gitignore b/.gitignore
  2. old mode 100644
  3. new mode 100755
  4. diff --git a/AUTHORS b/AUTHORS
  5. old mode 100644
  6. new mode 100755
  7. diff --git a/COPYING b/COPYING
  8. old mode 100644
  9. new mode 100755
  10. diff --git a/Makefile b/Makefile
  11. old mode 100644
  12. new mode 100755
  13. index 5831091..18f2d66
  14. --- a/Makefile
  15. +++ b/Makefile
  16. @@ -11,7 +11,8 @@ SRCS = common/mc.c common/predict.c common/pixel.c common/macroblock.c \
  17.         common/mvpred.c common/bitstream.c \
  18.         encoder/analyse.c encoder/me.c encoder/ratecontrol.c \
  19.         encoder/set.c encoder/macroblock.c encoder/cabac.c \
  20. -       encoder/cavlc.c encoder/encoder.c encoder/lookahead.c
  21. +       encoder/cavlc.c encoder/encoder.c encoder/lookahead.c \
  22. +       encoder/trellis.c
  23.  
  24.  SRCCLI = x264.c input/input.c input/timecode.c input/raw.c input/y4m.c \
  25.           output/raw.c output/matroska.c output/matroska_ebml.c \
  26. diff --git a/common/arm/asm.S b/common/arm/asm.S
  27. old mode 100644
  28. new mode 100755
  29. diff --git a/common/arm/cpu-a.S b/common/arm/cpu-a.S
  30. old mode 100644
  31. new mode 100755
  32. diff --git a/common/arm/dct-a.S b/common/arm/dct-a.S
  33. old mode 100644
  34. new mode 100755
  35. diff --git a/common/arm/dct.h b/common/arm/dct.h
  36. old mode 100644
  37. new mode 100755
  38. diff --git a/common/arm/deblock-a.S b/common/arm/deblock-a.S
  39. old mode 100644
  40. new mode 100755
  41. diff --git a/common/arm/mc-a.S b/common/arm/mc-a.S
  42. old mode 100644
  43. new mode 100755
  44. diff --git a/common/arm/mc-c.c b/common/arm/mc-c.c
  45. old mode 100644
  46. new mode 100755
  47. diff --git a/common/arm/mc.h b/common/arm/mc.h
  48. old mode 100644
  49. new mode 100755
  50. diff --git a/common/arm/pixel-a.S b/common/arm/pixel-a.S
  51. old mode 100644
  52. new mode 100755
  53. diff --git a/common/arm/pixel.h b/common/arm/pixel.h
  54. old mode 100644
  55. new mode 100755
  56. diff --git a/common/arm/predict-a.S b/common/arm/predict-a.S
  57. old mode 100644
  58. new mode 100755
  59. diff --git a/common/arm/predict-c.c b/common/arm/predict-c.c
  60. old mode 100644
  61. new mode 100755
  62. diff --git a/common/arm/predict.h b/common/arm/predict.h
  63. old mode 100644
  64. new mode 100755
  65. diff --git a/common/arm/quant-a.S b/common/arm/quant-a.S
  66. old mode 100644
  67. new mode 100755
  68. diff --git a/common/arm/quant.h b/common/arm/quant.h
  69. old mode 100644
  70. new mode 100755
  71. diff --git a/common/bitstream.c b/common/bitstream.c
  72. old mode 100644
  73. new mode 100755
  74. diff --git a/common/bitstream.h b/common/bitstream.h
  75. old mode 100644
  76. new mode 100755
  77. diff --git a/common/cabac.c b/common/cabac.c
  78. old mode 100644
  79. new mode 100755
  80. diff --git a/common/cabac.h b/common/cabac.h
  81. old mode 100644
  82. new mode 100755
  83. diff --git a/common/common.c b/common/common.c
  84. old mode 100644
  85. new mode 100755
  86. index a30d0d8..6818a01
  87. --- a/common/common.c
  88. +++ b/common/common.c
  89. @@ -976,6 +976,8 @@ int x264_param_parse( x264_param_t *p, const char *name, const char *value )
  90.          p->b_fake_interlaced = atobool(value);
  91.      OPT("frame-packing")
  92.          p->i_frame_packing = atoi(value);
  93. +    OPT("me-trellis")
  94. +        p->analyse.i_me_trellis = atoi(value);
  95.      else
  96.          return X264_PARAM_BAD_NAME;
  97.  #undef OPT
  98. diff --git a/common/common.h b/common/common.h
  99. old mode 100644
  100. new mode 100755
  101. index 7768571..fee8853
  102. --- a/common/common.h
  103. +++ b/common/common.h
  104. @@ -191,6 +191,7 @@ static const unsigned x264_scan8[16+2*4+3] =
  105.  #include "quant.h"
  106.  #include "cpu.h"
  107.  #include "threadpool.h"
  108. +#include "trellis.h"
  109.  
  110.  /****************************************************************************
  111.   * General functions
  112. @@ -771,6 +772,8 @@ struct x264_t
  113.              int8_t  topright_ref[2][3];
  114.          } cache;
  115.  
  116. +        me_trellis_t trellis;
  117. +
  118.          /* */
  119.          int     i_qp;       /* current qp */
  120.          int     i_chroma_qp;
  121. diff --git a/common/cpu.c b/common/cpu.c
  122. old mode 100644
  123. new mode 100755
  124. diff --git a/common/cpu.h b/common/cpu.h
  125. old mode 100644
  126. new mode 100755
  127. diff --git a/common/dct.c b/common/dct.c
  128. old mode 100644
  129. new mode 100755
  130. diff --git a/common/dct.h b/common/dct.h
  131. old mode 100644
  132. new mode 100755
  133. diff --git a/common/deblock.c b/common/deblock.c
  134. old mode 100644
  135. new mode 100755
  136. diff --git a/common/display-x11.c b/common/display-x11.c
  137. old mode 100644
  138. new mode 100755
  139. diff --git a/common/display.h b/common/display.h
  140. old mode 100644
  141. new mode 100755
  142. diff --git a/common/frame.c b/common/frame.c
  143. old mode 100644
  144. new mode 100755
  145. diff --git a/common/frame.h b/common/frame.h
  146. old mode 100644
  147. new mode 100755
  148. diff --git a/common/macroblock.c b/common/macroblock.c
  149. old mode 100644
  150. new mode 100755
  151. index 9eefc0a..61cbbb2
  152. --- a/common/macroblock.c
  153. +++ b/common/macroblock.c
  154. @@ -34,7 +34,7 @@ static NOINLINE void x264_mb_mc_0xywh( x264_t *h, int x, int y, int width, int h
  155.      int i_ref = h->mb.cache.ref[0][i8];
  156.      int mvx   = x264_clip3( h->mb.cache.mv[0][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] ) + 4*4*x;
  157.      int mvy   = x264_clip3( h->mb.cache.mv[0][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] ) + 4*4*y;
  158. -
  159. +    
  160.      h->mc.mc_luma( &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE,
  161.                     h->mb.pic.p_fref[0][i_ref], h->mb.pic.i_stride[0],
  162.                     mvx, mvy, 4*width, 4*height, &h->sh.weight[i_ref][0] );
  163. @@ -283,6 +283,34 @@ int x264_macroblock_cache_allocate( x264_t *h )
  164.              CHECKED_MALLOC( h->mb.p_weight_buf[i], luma_plane_size * sizeof(pixel) );
  165.      }
  166.  
  167. +    if ( h->param.analyse.i_me_trellis ) {
  168. +        h->mb.trellis.max_length = X264_MAX( h->mb.i_mb_width, h->mb.i_mb_height );
  169. +        h->mb.trellis.max_breadth = SQUARE2_CANDIDATES;
  170. +        h->mb.trellis.mb_stride_col = T_P_TOTAL+1;
  171. +        h->mb.trellis.mb_stride_row = h->mb.trellis.mb_stride_col * h->mb.i_mb_width;
  172. +        h->mb.trellis.t_stride = h->mb.trellis.max_breadth;
  173. +        h->mb.trellis.gen = 0;
  174. +                
  175. +        CHECKED_MALLOC( h->mb.trellis.t, h->mb.trellis.max_length * h->mb.trellis.max_breadth * sizeof( me_trellis_node_t ) );
  176. +        CHECKED_MALLOC( h->mb.trellis.mb, h->mb.i_mb_count * h->mb.trellis.mb_stride_col * sizeof( mb_t ) );
  177. +        CHECKED_MALLOC( h->mb.trellis.backup, h->mb.trellis.max_length * h->mb.trellis.mb_stride_col * sizeof( mb_t ) );
  178. +        
  179. +        CHECKED_MALLOCZERO( h->mb.trellis.cache_sad_gen[0], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(uint32_t) );
  180. +        CHECKED_MALLOCZERO( h->mb.trellis.cache_sad_gen[1], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(uint32_t) );
  181. +        CHECKED_MALLOCZERO( h->mb.trellis.cache_sad_tag[0], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(uint32_t) );
  182. +        CHECKED_MALLOCZERO( h->mb.trellis.cache_sad_tag[1], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(uint32_t) );
  183. +        CHECKED_MALLOC( h->mb.trellis.cache_sad_score[0], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(int) );
  184. +        CHECKED_MALLOC( h->mb.trellis.cache_sad_score[1], h->mb.i_mb_count * 9 * 8 * 8 * sizeof(int) );
  185. +        CHECKED_MALLOC( h->mb.trellis.cache_me_t[0], h->mb.i_mb_count * 9 * sizeof(x264_me_t) );
  186. +        CHECKED_MALLOC( h->mb.trellis.cache_me_t[1], h->mb.i_mb_count * 9 * sizeof(x264_me_t) );
  187. +        CHECKED_MALLOC( h->mb.trellis.cache_mode_intra, h->mb.i_mb_count * sizeof(int) );
  188. +        CHECKED_MALLOC( h->mb.trellis.cache_sad_intra, h->mb.i_mb_count * sizeof(int) );
  189. +        CHECKED_MALLOC( h->mb.trellis.best_type[0], h->mb.i_mb_count * sizeof(int) );
  190. +        CHECKED_MALLOC( h->mb.trellis.best_type[1], h->mb.i_mb_count * sizeof(int) );
  191. +        CHECKED_MALLOC( h->mb.trellis.best_part[0], h->mb.i_mb_count * sizeof(int) );
  192. +        CHECKED_MALLOC( h->mb.trellis.best_part[1], h->mb.i_mb_count * sizeof(int) );
  193. +    }
  194. +
  195.      return 0;
  196.  fail:
  197.      return -1;
  198. @@ -309,6 +337,27 @@ void x264_macroblock_cache_free( x264_t *h )
  199.      x264_free( h->mb.skipbp );
  200.      x264_free( h->mb.cbp );
  201.      x264_free( h->mb.qp );
  202. +
  203. +    if ( h->param.analyse.i_me_trellis ) {
  204. +        x264_free( h->mb.trellis.t );
  205. +        x264_free( h->mb.trellis.mb );
  206. +        x264_free( h->mb.trellis.backup );
  207. +        
  208. +        x264_free( h->mb.trellis.cache_sad_gen[0] );
  209. +        x264_free( h->mb.trellis.cache_sad_gen[1] );
  210. +        x264_free( h->mb.trellis.cache_sad_tag[0] );
  211. +        x264_free( h->mb.trellis.cache_sad_tag[1] );
  212. +        x264_free( h->mb.trellis.cache_sad_score[0] );
  213. +        x264_free( h->mb.trellis.cache_sad_score[1] );
  214. +        x264_free( h->mb.trellis.cache_me_t[0] );
  215. +        x264_free( h->mb.trellis.cache_me_t[1] );
  216. +        x264_free( h->mb.trellis.cache_mode_intra );
  217. +        x264_free( h->mb.trellis.cache_sad_intra );
  218. +        x264_free( h->mb.trellis.best_type[0] );
  219. +        x264_free( h->mb.trellis.best_type[1] );
  220. +        x264_free( h->mb.trellis.best_part[0] );
  221. +        x264_free( h->mb.trellis.best_part[1] );
  222. +    }
  223.  }
  224.  
  225.  int x264_macroblock_thread_allocate( x264_t *h, int b_lookahead )
  226. diff --git a/common/macroblock.h b/common/macroblock.h
  227. old mode 100644
  228. new mode 100755
  229. diff --git a/common/mc.c b/common/mc.c
  230. old mode 100644
  231. new mode 100755
  232. diff --git a/common/mc.h b/common/mc.h
  233. old mode 100644
  234. new mode 100755
  235. diff --git a/common/mvpred.c b/common/mvpred.c
  236. old mode 100644
  237. new mode 100755
  238. diff --git a/common/osdep.c b/common/osdep.c
  239. old mode 100644
  240. new mode 100755
  241. diff --git a/common/osdep.h b/common/osdep.h
  242. old mode 100644
  243. new mode 100755
  244. diff --git a/common/pixel.c b/common/pixel.c
  245. old mode 100644
  246. new mode 100755
  247. diff --git a/common/pixel.h b/common/pixel.h
  248. old mode 100644
  249. new mode 100755
  250. diff --git a/common/ppc/dct.c b/common/ppc/dct.c
  251. old mode 100644
  252. new mode 100755
  253. diff --git a/common/ppc/dct.h b/common/ppc/dct.h
  254. old mode 100644
  255. new mode 100755
  256. diff --git a/common/ppc/deblock.c b/common/ppc/deblock.c
  257. old mode 100644
  258. new mode 100755
  259. diff --git a/common/ppc/mc.c b/common/ppc/mc.c
  260. old mode 100644
  261. new mode 100755
  262. diff --git a/common/ppc/mc.h b/common/ppc/mc.h
  263. old mode 100644
  264. new mode 100755
  265. diff --git a/common/ppc/pixel.c b/common/ppc/pixel.c
  266. old mode 100644
  267. new mode 100755
  268. diff --git a/common/ppc/pixel.h b/common/ppc/pixel.h
  269. old mode 100644
  270. new mode 100755
  271. diff --git a/common/ppc/ppccommon.h b/common/ppc/ppccommon.h
  272. old mode 100644
  273. new mode 100755
  274. diff --git a/common/ppc/predict.c b/common/ppc/predict.c
  275. old mode 100644
  276. new mode 100755
  277. diff --git a/common/ppc/predict.h b/common/ppc/predict.h
  278. old mode 100644
  279. new mode 100755
  280. diff --git a/common/ppc/quant.c b/common/ppc/quant.c
  281. old mode 100644
  282. new mode 100755
  283. diff --git a/common/ppc/quant.h b/common/ppc/quant.h
  284. old mode 100644
  285. new mode 100755
  286. diff --git a/common/predict.c b/common/predict.c
  287. old mode 100644
  288. new mode 100755
  289. diff --git a/common/predict.h b/common/predict.h
  290. old mode 100644
  291. new mode 100755
  292. diff --git a/common/quant.c b/common/quant.c
  293. old mode 100644
  294. new mode 100755
  295. diff --git a/common/quant.h b/common/quant.h
  296. old mode 100644
  297. new mode 100755
  298. diff --git a/common/rectangle.c b/common/rectangle.c
  299. old mode 100644
  300. new mode 100755
  301. diff --git a/common/rectangle.h b/common/rectangle.h
  302. old mode 100644
  303. new mode 100755
  304. diff --git a/common/set.c b/common/set.c
  305. old mode 100644
  306. new mode 100755
  307. diff --git a/common/set.h b/common/set.h
  308. old mode 100644
  309. new mode 100755
  310. diff --git a/common/sparc/pixel.asm b/common/sparc/pixel.asm
  311. old mode 100644
  312. new mode 100755
  313. diff --git a/common/sparc/pixel.h b/common/sparc/pixel.h
  314. old mode 100644
  315. new mode 100755
  316. diff --git a/common/threadpool.c b/common/threadpool.c
  317. old mode 100644
  318. new mode 100755
  319. diff --git a/common/threadpool.h b/common/threadpool.h
  320. old mode 100644
  321. new mode 100755
  322. diff --git a/common/trellis.h b/common/trellis.h
  323. new file mode 100755
  324. index 0000000..e807e35
  325. --- /dev/null
  326. +++ b/common/trellis.h
  327. @@ -0,0 +1,179 @@
  328. +#ifndef TRELLIS_H
  329. +#define TRELLIS_H
  330. +
  331. +#define QP_FAKE     24
  332. +#define ME_RANGE    h->param.analyse.i_me_range
  333. +
  334. +#include "common.h"
  335. +#include "encoder/me.h"
  336. +
  337. +// tellis mode idents
  338. +enum trellis_mode_t {
  339. +    MODE_NONE               = 0,
  340. +    MODE_COMBINED_DIAMOND   = 1,
  341. +    MODE_DECIDE             = 2,
  342. +    MODE_DIAMOND            = 3,
  343. +    MODE_SQUARE2            = 4
  344. +};
  345. +
  346. +// trellis direction
  347. +enum trellis_direction_t {
  348. +    ROW_TRELLIS = 0,
  349. +    COL_TRELLIS = 1
  350. +};
  351. +
  352. +// trellis mb type identifiers
  353. +enum trellis_p_mb_type_t {
  354. +    T_P_16x16  = 0,
  355. +    T_P_8x8    = 1,
  356. +    T_P_16x8   = 2,
  357. +    T_P_8x16   = 3,
  358. +    T_P_INTRA  = 4,
  359. +    
  360. +    T_P_BEST   = 5,
  361. +    T_P_TOTAL  = 5
  362. +};
  363. +
  364. +// mv direction total
  365. +enum mv_candidates_t {
  366. +    COMBINED_DIAMOND_CANDIDATES = 21,
  367. +    DECIDE_CANDIDATES           = 5,
  368. +    DIAMOND_CANDIDATES          = 5,
  369. +    SQUARE2_CANDIDATES          = 25
  370. +};
  371. +
  372. +// mv direction indices
  373. +enum mv_candidate_index_t {
  374. +    T___    = 0,
  375. +    B___    = 1,
  376. +    L___    = 2,
  377. +    R___    = 3,
  378. +    C___    = 4,
  379. +    TL__    = 5,
  380. +    TR__    = 6,
  381. +    BL__    = 7,
  382. +    BR__    = 8,
  383. +    LL__    = 9,
  384. +    RR__    = 10,
  385. +    TT__    = 11,
  386. +    BB__    = 12,
  387. +    TTL_    = 13,
  388. +    TTLL    = 14,
  389. +    TTR_    = 15,
  390. +    TTRR    = 16,
  391. +    TLL_    = 17,
  392. +    TRR_    = 18,
  393. +    BBL_    = 19,
  394. +    BBLL    = 20,
  395. +    BBR_    = 21,
  396. +    BBRR    = 22,
  397. +    BLL_    = 23,
  398. +    BRR_    = 24
  399. +};
  400. +
  401. +// useful comparison
  402. +#define X264_MED(a,b,c) ( (a<=b) ? ( (c<=a) ? a : ( (c<=b) ? c : b ) ) : ( (c>=a) ? a : ( (c>=b) ? c : b ) ) )
  403. +
  404. +// copy memory between two records
  405. +#define CP_MV(a,b)    memcpy( a, b, 2*sizeof( int16_t ) )
  406. +#define CP_MB(a,b)    memcpy( a, b, sizeof( mb_t ) )
  407. +
  408. +// cost macros
  409. +
  410. +#define T_SAD_SCORE( class, partition, i_mv, row, col, x, y )\
  411. +    ((class==P_8x8) ? (h->mb.trellis.sad_p8x8[i_mv][row*h->mb.i_mb_stride+col][(2*ME_RANGE*(y+ME_RANGE))+(x+ME_RANGE)]) : \
  412. +    (((class==P_L0)&&(partition==D_16x16)) ? (h->mb.trellis.sad_p16x16[row*h->mb.i_mb_stride+col][(2*ME_RANGE*(y+ME_RANGE))+(x+ME_RANGE)]) : \
  413. +    (((class==P_L0)&&(partition==D_16x8)) ? (h->mb.trellis.sad_p16x8[i_mv][row*h->mb.i_mb_stride+col][((2*ME_RANGE)*(y+ME_RANGE))+(x+ME_RANGE)]) : \
  414. +    (((class==P_L0)&&(partition==D_8x16)) ? (h->mb.trellis.sad_p8x16[i_mv][row*h->mb.i_mb_stride+col][((2*ME_RANGE)*(y+ME_RANGE))+(x+ME_RANGE)]) : \
  415. +    (h->mb.trellis.sad_intra[row*h->mb.i_mb_stride+col])))))
  416. +
  417. +#define T_BITS_MVD( mx, my )\
  418. +    (h->cost_mv[QP_FAKE][(mx)<<2] + h->cost_mv[QP_FAKE][(my)<<2])
  419. +
  420. +#define T_BITS_MODE( class, partition )\
  421. +    (x264_lambda_tab[QP_FAKE] * ( ( IS_INTRA( class ) ) ? ( 0 ) : ( ( partition == D_16x16 ) ? ( 4 ) : ( ( partition == D_8x8 ) ? ( 12 ) : ( 6 ) ) ) ))
  422. +
  423. +#define T_COST_SAD( mx, my )\
  424. +    h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, &p_fref_w[(my)*stride+(mx)], stride )
  425. +    
  426. +// map x264 mb types to trellis mb types
  427. +
  428. +#define TRELLIS_TYPE( block )\
  429. +    ( IS_INTRA( block.class ) ? ( T_P_INTRA ) : \
  430. +    ( ( block.class == P_8x8 ) ? ( T_P_8x8 ) : \
  431. +    ( ( block.part[0] = D_16x16 ) ? ( T_P_16x16 ) : \
  432. +    ( ( block.part[0] = D_16x8 ) ? ( T_P_16x8 ) : ( T_P_8x16 ) \
  433. +    ) ) ) )
  434. +
  435. +// holds mvs for partitioned block.  unused mvs are zero by default.
  436. +typedef struct mb_t {
  437. +    int class;
  438. +    int costSAD;
  439. +    int costMV;
  440. +    int16_t mv_min[2];
  441. +    int16_t mv_max[2];
  442. +    int8_t part[4];
  443. +    int16_t ((mv[2])[4])[2];
  444. +    uint8_t (ref[4])[2];
  445. +} mb_t;
  446. +
  447. +typedef struct me_trellis_node_t {
  448. +    mb_t block;
  449. +    int8_t parent;
  450. +    int accCost;
  451. +    int costCC;
  452. +    int costDC;
  453. +    int costDL;
  454. +    int costDR;
  455. +    int costCR;
  456. +} me_trellis_node_t;
  457. +
  458. +
  459. +typedef struct me_trellis_t {
  460. +    me_trellis_node_t *t;
  461. +    mb_t *mb;
  462. +    mb_t *backup;
  463. +
  464. +    int mode;
  465. +    int direction;
  466. +    int breadth;
  467. +    int max_breadth;
  468. +    int length;
  469. +    int max_length;
  470. +    uint32_t gen;
  471. +    
  472. +    int b_collected;
  473. +    
  474. +    int mb_stride_col;
  475. +    int mb_stride_row;
  476. +    int t_stride;
  477. +    
  478. +    uint32_t (*cache_sad_gen[2])[9][8][8];
  479. +    uint32_t (*cache_sad_tag[2])[9][8][8];
  480. +    int (*cache_sad_score[2])[9][8][8];
  481. +    x264_me_t (*cache_me_t[2])[9];
  482. +    int *cache_mode_intra;
  483. +    int *cache_sad_intra;
  484. +    int (*best_type[2]);
  485. +    int (*best_part[2]);
  486. +} me_trellis_t;
  487. +
  488. +int get_cached_fpel_sad( x264_t *h, int class, int partition, int i_list, int i_mv, int row, int col, int x, int y );
  489. +void fill_me_trellis_params( x264_t *h, int mode, int direction );
  490. +void get_left_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv );
  491. +void get_top_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv );
  492. +void get_top_right_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv );
  493. +void get_top_left_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv );
  494. +void choose_predictors( x264_t *h, mb_t *pred, int row, int col, mb_t *current, mb_t *left, mb_t *top, mb_t *diag );
  495. +int cost_mb( x264_t *h, int row, int col, int i_mv_max, mb_t *block, mb_t *left, mb_t *top, mb_t *diag );
  496. +void collect_candidates( x264_t *h );
  497. +int sad_cost( x264_t *h );
  498. +int mv_cost( x264_t *h );
  499. +void fill_trellis( x264_t *h, int i_frame );
  500. +void adjust_trellis( x264_t *h, int sub );
  501. +void generate_alternate_mvs( x264_t *h, int16_t mv[SQUARE2_CANDIDATES][2], mb_t *block, int i_list, int i_mv );
  502. +void run_row_trellis( x264_t *h, int iter );
  503. +void scale_trellis_mv( x264_t *h );
  504. +void write_back_trellis_mv( x264_t *h, int i_mb );
  505. +
  506. +#endif
  507. diff --git a/common/visualize.c b/common/visualize.c
  508. old mode 100644
  509. new mode 100755
  510. diff --git a/common/visualize.h b/common/visualize.h
  511. old mode 100644
  512. new mode 100755
  513. diff --git a/common/vlc.c b/common/vlc.c
  514. old mode 100644
  515. new mode 100755
  516. diff --git a/common/win32thread.c b/common/win32thread.c
  517. old mode 100644
  518. new mode 100755
  519. diff --git a/common/win32thread.h b/common/win32thread.h
  520. old mode 100644
  521. new mode 100755
  522. diff --git a/common/x86/bitstream-a.asm b/common/x86/bitstream-a.asm
  523. old mode 100644
  524. new mode 100755
  525. diff --git a/common/x86/cabac-a.asm b/common/x86/cabac-a.asm
  526. old mode 100644
  527. new mode 100755
  528. diff --git a/common/x86/const-a.asm b/common/x86/const-a.asm
  529. old mode 100644
  530. new mode 100755
  531. diff --git a/common/x86/cpu-a.asm b/common/x86/cpu-a.asm
  532. old mode 100644
  533. new mode 100755
  534. diff --git a/common/x86/dct-32.asm b/common/x86/dct-32.asm
  535. old mode 100644
  536. new mode 100755
  537. diff --git a/common/x86/dct-64.asm b/common/x86/dct-64.asm
  538. old mode 100644
  539. new mode 100755
  540. diff --git a/common/x86/dct-a.asm b/common/x86/dct-a.asm
  541. old mode 100644
  542. new mode 100755
  543. diff --git a/common/x86/dct.h b/common/x86/dct.h
  544. old mode 100644
  545. new mode 100755
  546. diff --git a/common/x86/deblock-a.asm b/common/x86/deblock-a.asm
  547. old mode 100644
  548. new mode 100755
  549. diff --git a/common/x86/mc-a.asm b/common/x86/mc-a.asm
  550. old mode 100644
  551. new mode 100755
  552. diff --git a/common/x86/mc-a2.asm b/common/x86/mc-a2.asm
  553. old mode 100644
  554. new mode 100755
  555. diff --git a/common/x86/mc-c.c b/common/x86/mc-c.c
  556. old mode 100644
  557. new mode 100755
  558. diff --git a/common/x86/mc.h b/common/x86/mc.h
  559. old mode 100644
  560. new mode 100755
  561. diff --git a/common/x86/pixel-32.asm b/common/x86/pixel-32.asm
  562. old mode 100644
  563. new mode 100755
  564. diff --git a/common/x86/pixel-a.asm b/common/x86/pixel-a.asm
  565. old mode 100644
  566. new mode 100755
  567. diff --git a/common/x86/pixel.h b/common/x86/pixel.h
  568. old mode 100644
  569. new mode 100755
  570. diff --git a/common/x86/predict-a.asm b/common/x86/predict-a.asm
  571. old mode 100644
  572. new mode 100755
  573. diff --git a/common/x86/predict-c.c b/common/x86/predict-c.c
  574. old mode 100644
  575. new mode 100755
  576. diff --git a/common/x86/predict.h b/common/x86/predict.h
  577. old mode 100644
  578. new mode 100755
  579. diff --git a/common/x86/quant-a.asm b/common/x86/quant-a.asm
  580. old mode 100644
  581. new mode 100755
  582. diff --git a/common/x86/quant.h b/common/x86/quant.h
  583. old mode 100644
  584. new mode 100755
  585. diff --git a/common/x86/sad-a.asm b/common/x86/sad-a.asm
  586. old mode 100644
  587. new mode 100755
  588. diff --git a/common/x86/sad16-a.asm b/common/x86/sad16-a.asm
  589. old mode 100644
  590. new mode 100755
  591. diff --git a/common/x86/util.h b/common/x86/util.h
  592. old mode 100644
  593. new mode 100755
  594. diff --git a/common/x86/x86inc.asm b/common/x86/x86inc.asm
  595. old mode 100644
  596. new mode 100755
  597. diff --git a/common/x86/x86util.asm b/common/x86/x86util.asm
  598. old mode 100644
  599. new mode 100755
  600. diff --git a/configure b/configure
  601. index d164414..b00900f 100755
  602. --- a/configure
  603. +++ b/configure
  604. @@ -857,7 +857,7 @@ if [ "$strip" = "yes" ]; then
  605.  fi
  606.  
  607.  if [ "$debug" = "yes" ]; then
  608. -    CFLAGS="-O1 -g $CFLAGS"
  609. +    CFLAGS="-O0 -g $CFLAGS"
  610.  elif [ $ARCH = ARM ]; then
  611.      # arm-gcc-4.2 produces incorrect output with -ffast-math
  612.      # and it doesn't save any speed anyway on 4.4, so disable it
  613. diff --git a/doc/ratecontrol.txt b/doc/ratecontrol.txt
  614. old mode 100644
  615. new mode 100755
  616. diff --git a/doc/regression_test.txt b/doc/regression_test.txt
  617. old mode 100644
  618. new mode 100755
  619. diff --git a/doc/standards.txt b/doc/standards.txt
  620. old mode 100644
  621. new mode 100755
  622. diff --git a/doc/threads.txt b/doc/threads.txt
  623. old mode 100644
  624. new mode 100755
  625. diff --git a/doc/vui.txt b/doc/vui.txt
  626. old mode 100644
  627. new mode 100755
  628. diff --git a/encoder/analyse.c b/encoder/analyse.c
  629. old mode 100644
  630. new mode 100755
  631. index 5b5083d..96061b7
  632. --- a/encoder/analyse.c
  633. +++ b/encoder/analyse.c
  634. @@ -809,7 +809,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  635.          /* cavlc mb type prefix */
  636.          a->i_satd_i16x16 += lambda * i_mb_b_cost_table[I_16x16];
  637.  
  638. -    if( a->i_satd_i16x16 > i16x16_thresh )
  639. +    if( !h->param.analyse.i_me_trellis && a->i_satd_i16x16 > i16x16_thresh )
  640.          return;
  641.  
  642.      /* 8x8 prediction selection */
  643. @@ -911,7 +911,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  644.          }
  645.          /* Not heavily tuned */
  646.          static const uint8_t i8x8_thresh[11] = { 4, 4, 4, 5, 5, 5, 6, 6, 6, 6, 6 };
  647. -        if( X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  648. +        if( !h->param.analyse.i_me_trellis && X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  649.              return;
  650.      }
  651.  
  652. @@ -1258,6 +1258,12 @@ static void x264_mb_analyse_inter_p16x16( x264_t *h, x264_mb_analysis_t *a )
  653.      /* 16x16 Search on all ref frame */
  654.      m.i_pixel = PIXEL_16x16;
  655.      LOAD_FENC( &m, h->mb.pic.p_fenc, 0, 0 );
  656. +    
  657. +    if ( h->param.analyse.i_me_trellis ) {
  658. +        m.cache_sad_gen = &(h->mb.trellis.cache_sad_gen[0][h->mb.i_mb_xy][0]);
  659. +        m.cache_sad_tag = &(h->mb.trellis.cache_sad_tag[0][h->mb.i_mb_xy][0]);
  660. +        m.cache_sad_score = &(h->mb.trellis.cache_sad_score[0][h->mb.i_mb_xy][0]);
  661. +    }
  662.  
  663.      a->l0.me16x16.cost = INT_MAX;
  664.      for( int i_ref = 0; i_ref < h->mb.pic.i_fref[0]; i_ref++ )
  665. @@ -1307,6 +1313,9 @@ static void x264_mb_analyse_inter_p16x16( x264_t *h, x264_mb_analysis_t *a )
  666.          if( m.cost < a->l0.me16x16.cost )
  667.              h->mc.memcpy_aligned( &a->l0.me16x16, &m, sizeof(x264_me_t) );
  668.      }
  669. +    
  670. +    if ( h->param.analyse.i_me_trellis )
  671. +        h->mc.memcpy_aligned( &(h->mb.trellis.cache_me_t[0][h->mb.i_mb_xy][0]), &(a->l0.me16x16), sizeof(x264_me_t) );
  672.  
  673.      x264_macroblock_cache_ref( h, 0, 0, 4, 4, 0, a->l0.me16x16.i_ref );
  674.      assert( a->l0.me16x16.mv[1] <= h->mb.mv_max_spel[1] || h->i_thread_frames == 1 );
  675. @@ -1440,6 +1449,12 @@ static void x264_mb_analyse_inter_p8x8( x264_t *h, x264_mb_analysis_t *a )
  676.          x264_me_t *m = &a->l0.me8x8[i];
  677.          int x8 = i&1;
  678.          int y8 = i>>1;
  679. +        
  680. +        if ( h->param.analyse.i_me_trellis ) {
  681. +            m->cache_sad_gen = &(h->mb.trellis.cache_sad_gen[0][h->mb.i_mb_xy][1+i]);
  682. +            m->cache_sad_tag = &(h->mb.trellis.cache_sad_tag[0][h->mb.i_mb_xy][1+i]);
  683. +            m->cache_sad_score = &(h->mb.trellis.cache_sad_score[0][h->mb.i_mb_xy][1+i]);
  684. +        }
  685.  
  686.          m->i_pixel = PIXEL_8x8;
  687.          m->i_ref_cost = i_ref_cost;
  688. @@ -1462,6 +1477,10 @@ static void x264_mb_analyse_inter_p8x8( x264_t *h, x264_mb_analysis_t *a )
  689.          m->cost += i_ref_cost;
  690.          if( !h->param.b_cabac || (h->param.analyse.inter & X264_ANALYSE_PSUB8x8) )
  691.              m->cost += a->i_lambda * i_sub_mb_p_cost_table[D_L0_8x8];
  692. +            
  693. +        if ( h->param.analyse.i_me_trellis )
  694. +            h->mc.memcpy_aligned( &(h->mb.trellis.cache_me_t[0][h->mb.i_mb_xy][1+i]), m, sizeof(x264_me_t) );
  695. +
  696.      }
  697.  
  698.      a->l0.i_cost8x8 = a->l0.me8x8[0].cost + a->l0.me8x8[1].cost +
  699. @@ -1493,6 +1512,12 @@ static void x264_mb_analyse_inter_p16x8( x264_t *h, x264_mb_analysis_t *a, int i
  700.  
  701.          m.i_pixel = PIXEL_16x8;
  702.  
  703. +        if ( h->param.analyse.i_me_trellis ) {
  704. +            m.cache_sad_gen = &(h->mb.trellis.cache_sad_gen[0][h->mb.i_mb_xy][5+i]);
  705. +            m.cache_sad_tag = &(h->mb.trellis.cache_sad_tag[0][h->mb.i_mb_xy][5+i]);
  706. +            m.cache_sad_score = &(h->mb.trellis.cache_sad_score[0][h->mb.i_mb_xy][5+i]);
  707. +        }
  708. +
  709.          LOAD_FENC( &m, p_fenc, 0, 8*i );
  710.          l0m->cost = INT_MAX;
  711.          for( int j = 0; j < i_ref8s; j++ )
  712. @@ -1524,10 +1549,13 @@ static void x264_mb_analyse_inter_p16x8( x264_t *h, x264_mb_analysis_t *a, int i
  713.              if( m.cost < l0m->cost )
  714.                  h->mc.memcpy_aligned( l0m, &m, sizeof(x264_me_t) );
  715.          }
  716. +        
  717. +        if ( h->param.analyse.i_me_trellis )
  718. +            h->mc.memcpy_aligned( &(h->mb.trellis.cache_me_t[0][h->mb.i_mb_xy][5+i]), l0m, sizeof(x264_me_t) );
  719.  
  720.          /* Early termination based on the current SATD score of partition[0]
  721.             plus the estimated SATD score of partition[1] */
  722. -        if( !i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  723. +        if( !(h->param.analyse.i_me_trellis) && !i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  724.          {
  725.              a->l0.i_cost16x8 = COST_MAX;
  726.              return;
  727. @@ -1559,6 +1587,12 @@ static void x264_mb_analyse_inter_p8x16( x264_t *h, x264_mb_analysis_t *a, int i
  728.  
  729.          m.i_pixel = PIXEL_8x16;
  730.  
  731. +        if ( h->param.analyse.i_me_trellis ) {
  732. +            m.cache_sad_gen = &(h->mb.trellis.cache_sad_gen[0][h->mb.i_mb_xy][7+i]);
  733. +            m.cache_sad_tag = &(h->mb.trellis.cache_sad_tag[0][h->mb.i_mb_xy][7+i]);
  734. +            m.cache_sad_score = &(h->mb.trellis.cache_sad_score[0][h->mb.i_mb_xy][7+i]);
  735. +        }
  736. +
  737.          LOAD_FENC( &m, p_fenc, 8*i, 0 );
  738.          l0m->cost = INT_MAX;
  739.          for( int j = 0; j < i_ref8s; j++ )
  740. @@ -1589,10 +1623,13 @@ static void x264_mb_analyse_inter_p8x16( x264_t *h, x264_mb_analysis_t *a, int i
  741.              if( m.cost < l0m->cost )
  742.                  h->mc.memcpy_aligned( l0m, &m, sizeof(x264_me_t) );
  743.          }
  744. +        
  745. +        if ( h->param.analyse.i_me_trellis )
  746. +            h->mc.memcpy_aligned( &(h->mb.trellis.cache_me_t[0][h->mb.i_mb_xy][7+i]), l0m, sizeof(x264_me_t) );
  747.  
  748.          /* Early termination based on the current SATD score of partition[0]
  749.             plus the estimated SATD score of partition[1] */
  750. -        if( !i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  751. +        if( !h->param.analyse.i_me_trellis && !i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  752.          {
  753.              a->l0.i_cost8x16 = COST_MAX;
  754.              return;
  755. @@ -2920,8 +2957,8 @@ intra_analysis:
  756.              i_partition = D_16x16;
  757.              i_cost = analysis.l0.me16x16.cost;
  758.  
  759. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  760. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost )
  761. +            if( ( ( h->param.analyse.i_me_trellis || ( flags & X264_ANALYSE_PSUB16x16 ) ) &&
  762. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost ) )
  763.              {
  764.                  i_type = P_8x8;
  765.                  i_partition = D_8x8;
  766. @@ -2956,8 +2993,8 @@ intra_analysis:
  767.  
  768.              /* Now do 16x8/8x16 */
  769.              i_thresh16x8 = analysis.l0.me8x8[1].cost_mv + analysis.l0.me8x8[2].cost_mv;
  770. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  771. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8 )
  772. +            if( h->param.analyse.i_me_trellis || ( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  773. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8 ) )
  774.              {
  775.                  int i_avg_mv_ref_cost = (analysis.l0.me8x8[2].cost_mv + analysis.l0.me8x8[2].i_ref_cost
  776.                                        + analysis.l0.me8x8[3].cost_mv + analysis.l0.me8x8[3].i_ref_cost + 1) >> 1;
  777. @@ -2975,7 +3012,11 @@ intra_analysis:
  778.              }
  779.  
  780.              h->mb.i_partition = i_partition;
  781. -
  782. +            if ( h->param.analyse.i_me_trellis )
  783. +                h->mb.trellis.best_part[0][h->mb.i_mb_xy] = h->mb.i_partition;
  784. +                
  785. +            //printf("(%d,%d,%d,%d)\n",analysis.l0.me16x16.cost,analysis.l0.i_cost8x8,analysis.l0.i_cost16x8,analysis.l0.i_cost8x16);
  786. +                
  787.              /* refine qpel */
  788.              //FIXME mb_type costs?
  789.              if( analysis.i_mbrd || !h->mb.i_subpel_refine )
  790. @@ -3055,7 +3096,18 @@ intra_analysis:
  791.              i_satd_intra = X264_MIN3( analysis.i_satd_i16x16,
  792.                                        analysis.i_satd_i8x8,
  793.                                        analysis.i_satd_i4x4 );
  794. -
  795. +            
  796. +            
  797. +            if ( h->param.analyse.i_me_trellis ) {
  798. +                h->mb.trellis.cache_sad_intra[h->mb.i_mb_xy] = i_satd_intra;
  799. +                if ( i_satd_intra == analysis.i_satd_i16x16 )
  800. +                    h->mb.trellis.cache_mode_intra[h->mb.i_mb_xy] = I_16x16;
  801. +                else if ( i_satd_intra == analysis.i_satd_i8x8 )
  802. +                    h->mb.trellis.cache_mode_intra[h->mb.i_mb_xy] = I_8x8;
  803. +                else if ( i_satd_intra == analysis.i_satd_i4x4 )
  804. +                    h->mb.trellis.cache_mode_intra[h->mb.i_mb_xy] = I_4x4;
  805. +            }
  806. +                                      
  807.              if( analysis.i_mbrd )
  808.              {
  809.                  x264_mb_analyse_p_rd( h, &analysis, X264_MIN(i_satd_inter, i_satd_intra) );
  810. @@ -3078,6 +3130,10 @@ intra_analysis:
  811.              COPY2_IF_LT( i_cost, analysis.i_satd_pcm, i_type, I_PCM );
  812.  
  813.              h->mb.i_type = i_type;
  814. +            if ( h->param.analyse.i_me_trellis )
  815. +                h->mb.trellis.best_type[0][h->mb.i_mb_xy] = h->mb.i_type;
  816. +
  817. +            //printf("%d,%d\n",i_type,i_partition);
  818.  
  819.              if( analysis.b_force_intra && !IS_INTRA(i_type) )
  820.              {
  821. diff --git a/encoder/analyse.h b/encoder/analyse.h
  822. old mode 100644
  823. new mode 100755
  824. diff --git a/encoder/cabac.c b/encoder/cabac.c
  825. old mode 100644
  826. new mode 100755
  827. diff --git a/encoder/cavlc.c b/encoder/cavlc.c
  828. old mode 100644
  829. new mode 100755
  830. diff --git a/encoder/encoder.c b/encoder/encoder.c
  831. old mode 100644
  832. new mode 100755
  833. index 3571a91..ae8a39a
  834. --- a/encoder/encoder.c
  835. +++ b/encoder/encoder.c
  836. @@ -1762,7 +1762,7 @@ static void x264_fdec_filter_row( x264_t *h, int mb_y, int b_inloop )
  837.      if( min_y < h->i_threadslice_start )
  838.          return;
  839.  
  840. -    if( b_deblock )
  841. +    if( b_deblock && !( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected ) )
  842.          for( int y = min_y; y < mb_y; y += (1 << SLICE_MBAFF) )
  843.              x264_frame_deblock_row( h, y );
  844.  
  845. @@ -1776,7 +1776,7 @@ static void x264_fdec_filter_row( x264_t *h, int mb_y, int b_inloop )
  846.                          h->fdec->plane[p]     + i*h->fdec->i_stride[p],
  847.                          h->mb.i_mb_width*16*sizeof(pixel) );
  848.  
  849. -    if( b_hpel )
  850. +    if( b_hpel && !( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected ) )
  851.      {
  852.          int end = mb_y == h->mb.i_mb_height;
  853.          x264_frame_expand_border( h, h->fdec, min_y, end );
  854. @@ -1797,7 +1797,7 @@ static void x264_fdec_filter_row( x264_t *h, int mb_y, int b_inloop )
  855.      if( h->i_thread_frames > 1 && h->fdec->b_kept_as_ref )
  856.          x264_frame_cond_broadcast( h->fdec, mb_y*16 + (b_end ? 10000 : -(X264_THREAD_HEIGHT << SLICE_MBAFF)) );
  857.  
  858. -    if( b_measure_quality )
  859. +    if( b_measure_quality && !( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected ) )
  860.      {
  861.          maxpix_y = X264_MIN( maxpix_y, h->param.i_height );
  862.          if( h->param.analyse.b_psnr )
  863. @@ -2023,9 +2023,14 @@ static int x264_slice_write( x264_t *h )
  864.      i_mb_x = h->sh.i_first_mb % h->mb.i_mb_width;
  865.      i_skip = 0;
  866.      int mb_size[2];
  867. -
  868. +    
  869. +    h->param.analyse.b_cache_sads = ( h->param.analyse.i_me_trellis && ! h->mb.trellis.b_collected );
  870. +    if ( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected )
  871. +        h->mb.trellis.gen++; // update cache generation
  872. +    
  873.      while( 1 )
  874.      {
  875. +
  876.          mb_xy = i_mb_x + i_mb_y * h->mb.i_mb_width;
  877.          int mb_spos = bs_pos(&h->out.bs) + x264_cabac_pos(&h->cabac);
  878.  
  879. @@ -2079,7 +2084,10 @@ static int x264_slice_write( x264_t *h )
  880.              x264_macroblock_cache_load_progressive( h, i_mb_x, i_mb_y );
  881.  
  882.          x264_macroblock_analyse( h );
  883. -
  884. +        
  885. +        if ( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && h->mb.trellis.b_collected )
  886. +            write_back_trellis_mv( h, mb_xy );
  887. +            
  888.          /* encode this macroblock -> be careful it can change the mb type to P_SKIP if needed */
  889.  reencode:
  890.          x264_macroblock_encode( h );
  891. @@ -2190,59 +2198,62 @@ reencode:
  892.          /* save cache */
  893.          x264_macroblock_cache_save( h );
  894.  
  895. -        /* accumulate mb stats */
  896. -        h->stat.frame.i_mb_count[h->mb.i_type]++;
  897. +        if( !( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected ) ) {
  898. +            /* accumulate mb stats */
  899. +            h->stat.frame.i_mb_count[h->mb.i_type]++;
  900.  
  901. -        int b_intra = IS_INTRA( h->mb.i_type );
  902. -        int b_skip = IS_SKIP( h->mb.i_type );
  903. -        if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
  904. -        {
  905. -            if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
  906. +            int b_intra = IS_INTRA( h->mb.i_type );
  907. +            int b_skip = IS_SKIP( h->mb.i_type );
  908. +            if( h->param.i_log_level >= X264_LOG_INFO || h->param.rc.b_stat_write )
  909.              {
  910. -                if( h->mb.i_partition != D_8x8 )
  911. -                        h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
  912. -                    else
  913. -                        for( int i = 0; i < 4; i++ )
  914. -                            h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
  915. -                if( h->param.i_frame_reference > 1 )
  916. -                    for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
  917. -                        for( int i = 0; i < 4; i++ )
  918. -                        {
  919. -                            int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
  920. -                            if( i_ref >= 0 )
  921. -                                h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
  922. -                        }
  923. +                if( !b_intra && !b_skip && !IS_DIRECT( h->mb.i_type ) )
  924. +                {
  925. +                    if( h->mb.i_partition != D_8x8 )
  926. +                            h->stat.frame.i_mb_partition[h->mb.i_partition] += 4;
  927. +                        else
  928. +                            for( int i = 0; i < 4; i++ )
  929. +                                h->stat.frame.i_mb_partition[h->mb.i_sub_partition[i]] ++;
  930. +                    if( h->param.i_frame_reference > 1 )
  931. +                        for( int i_list = 0; i_list <= (h->sh.i_type == SLICE_TYPE_B); i_list++ )
  932. +                            for( int i = 0; i < 4; i++ )
  933. +                            {
  934. +                                int i_ref = h->mb.cache.ref[i_list][ x264_scan8[4*i] ];
  935. +                                if( i_ref >= 0 )
  936. +                                    h->stat.frame.i_mb_count_ref[i_list][i_ref] ++;
  937. +                            }
  938. +                }
  939.              }
  940. -        }
  941.  
  942. -        if( h->param.i_log_level >= X264_LOG_INFO )
  943. -        {
  944. -            if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
  945. +            if( h->param.i_log_level >= X264_LOG_INFO )
  946.              {
  947. -                int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
  948. -                           + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
  949. -                h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
  950. -                h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
  951. -                h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
  952. -            }
  953. -            if( h->mb.i_cbp_luma && !b_intra )
  954. -            {
  955. -                h->stat.frame.i_mb_count_8x8dct[0] ++;
  956. -                h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
  957. -            }
  958. -            if( b_intra && h->mb.i_type != I_PCM )
  959. -            {
  960. -                if( h->mb.i_type == I_16x16 )
  961. -                    h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
  962. -                else if( h->mb.i_type == I_8x8 )
  963. -                    for( int i = 0; i < 16; i += 4 )
  964. -                        h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
  965. -                else //if( h->mb.i_type == I_4x4 )
  966. -                    for( int i = 0; i < 16; i++ )
  967. -                        h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
  968. -                h->stat.frame.i_mb_pred_mode[3][x264_mb_pred_mode8x8c_fix[h->mb.i_chroma_pred_mode]]++;
  969. +                if( h->mb.i_cbp_luma | h->mb.i_cbp_chroma )
  970. +                {
  971. +                    int cbpsum = (h->mb.i_cbp_luma&1) + ((h->mb.i_cbp_luma>>1)&1)
  972. +                               + ((h->mb.i_cbp_luma>>2)&1) + (h->mb.i_cbp_luma>>3);
  973. +                    h->stat.frame.i_mb_cbp[!b_intra + 0] += cbpsum;
  974. +                    h->stat.frame.i_mb_cbp[!b_intra + 2] += !!h->mb.i_cbp_chroma;
  975. +                    h->stat.frame.i_mb_cbp[!b_intra + 4] += h->mb.i_cbp_chroma >> 1;
  976. +                }
  977. +                if( h->mb.i_cbp_luma && !b_intra )
  978. +                {
  979. +                    h->stat.frame.i_mb_count_8x8dct[0] ++;
  980. +                    h->stat.frame.i_mb_count_8x8dct[1] += h->mb.b_transform_8x8;
  981. +                }
  982. +                if( b_intra && h->mb.i_type != I_PCM )
  983. +                {
  984. +                    if( h->mb.i_type == I_16x16 )
  985. +                        h->stat.frame.i_mb_pred_mode[0][h->mb.i_intra16x16_pred_mode]++;
  986. +                    else if( h->mb.i_type == I_8x8 )
  987. +                        for( int i = 0; i < 16; i += 4 )
  988. +                            h->stat.frame.i_mb_pred_mode[1][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
  989. +                    else //if( h->mb.i_type == I_4x4 )
  990. +                        for( int i = 0; i < 16; i++ )
  991. +                            h->stat.frame.i_mb_pred_mode[2][h->mb.cache.intra4x4_pred_mode[x264_scan8[i]]]++;
  992. +                    h->stat.frame.i_mb_pred_mode[3][x264_mb_pred_mode8x8c_fix[h->mb.i_chroma_pred_mode]]++;
  993. +                }
  994. +                h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
  995.              }
  996. -            h->stat.frame.i_mb_field[b_intra?0:b_skip?2:1] += MB_INTERLACED;
  997. +            
  998.          }
  999.  
  1000.          /* calculate deblock strength values (actual deblocking is done per-row along with hpel) */
  1001. @@ -2274,6 +2285,28 @@ reencode:
  1002.              i_mb_x = 0;
  1003.          }
  1004.      }
  1005. +
  1006. +    if ( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && !h->mb.trellis.b_collected ) {
  1007. +        collect_candidates( h );
  1008. +        h->mb.trellis.b_collected = 1;
  1009. +        if ( h->param.analyse.i_me_trellis == X264_ME_TRELLIS_COMBINED_DIA ) {
  1010. +            fill_me_trellis_params( h, MODE_COMBINED_DIAMOND, ROW_TRELLIS );
  1011. +            run_row_trellis( h, 3 );
  1012. +        }
  1013. +        else {
  1014. +            fill_me_trellis_params( h, MODE_DECIDE, ROW_TRELLIS );
  1015. +            run_row_trellis( h, 1 );
  1016. +            if ( h->param.analyse.i_me_trellis == X264_ME_TRELLIS_DIA )
  1017. +                fill_me_trellis_params( h, MODE_DIAMOND, ROW_TRELLIS );
  1018. +            else
  1019. +                fill_me_trellis_params( h, MODE_SQUARE2, ROW_TRELLIS );                
  1020. +            run_row_trellis( h, 3 );
  1021. +        }
  1022. +        scale_trellis_mv( h );
  1023. +        
  1024. +        return 0;
  1025. +    }
  1026. +    
  1027.      h->out.nal[h->out.i_nal].i_last_mb = h->sh.i_last_mb;
  1028.  
  1029.      if( h->param.b_cabac )
  1030. @@ -2300,6 +2333,9 @@ reencode:
  1031.                                    - h->stat.frame.i_mv_bits;
  1032.          x264_fdec_filter_row( h, h->i_threadslice_end, 1 );
  1033.      }
  1034. +    
  1035. +    if ( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I && h->mb.trellis.b_collected )
  1036. +        h->mb.trellis.b_collected = 0;
  1037.  
  1038.      return 0;
  1039.  }
  1040. @@ -2373,6 +2409,10 @@ static void *x264_slices_write( x264_t *h )
  1041.          h->sh.i_last_mb = X264_MIN( h->sh.i_last_mb, last_thread_mb );
  1042.          if( x264_stack_align( x264_slice_write, h ) )
  1043.              return (void *)-1;
  1044. +        if( h->param.analyse.i_me_trellis && h->sh.i_type != SLICE_TYPE_I ) {
  1045. +            if ( x264_stack_align( x264_slice_write, h ) )
  1046. +                return (void *)-1;
  1047. +        }
  1048.          h->sh.i_first_mb = h->sh.i_last_mb + 1;
  1049.          // if i_first_mb is not the last mb in a row then go to the next mb in MBAFF order
  1050.          if( SLICE_MBAFF && h->sh.i_first_mb % h->mb.i_mb_width )
  1051. @@ -2617,6 +2657,8 @@ int     x264_encoder_encode( x264_t *h,
  1052.      }
  1053.  
  1054.      h->i_frame++;
  1055. +    // trellis cache fix
  1056. +    h->param.analyse.b_cache_sads = 0;
  1057.      /* 3: The picture is analyzed in the lookahead */
  1058.      if( !h->frames.current[0] )
  1059.          x264_lookahead_get_frames( h );
  1060. diff --git a/encoder/lookahead.c b/encoder/lookahead.c
  1061. old mode 100644
  1062. new mode 100755
  1063. diff --git a/encoder/macroblock.c b/encoder/macroblock.c
  1064. old mode 100644
  1065. new mode 100755
  1066. diff --git a/encoder/macroblock.h b/encoder/macroblock.h
  1067. old mode 100644
  1068. new mode 100755
  1069. diff --git a/encoder/me.c b/encoder/me.c
  1070. old mode 100644
  1071. new mode 100755
  1072. index 305e0c4..5fe97cb
  1073. --- a/encoder/me.c
  1074. +++ b/encoder/me.c
  1075. @@ -61,9 +61,13 @@ static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_ite
  1076.  
  1077.  #define COST_MV( mx, my )\
  1078.  {\
  1079. -    int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE,\
  1080. -                   &p_fref_w[(my)*stride+(mx)], stride )\
  1081. -             + BITS_MVD(mx,my);\
  1082. +    int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, &p_fref_w[(my)*stride+(mx)], stride );\
  1083. +    if ( h->param.analyse.b_cache_sads ) {\
  1084. +        (*(m->cache_sad_gen))[(my)&7][(mx)&7] = h->mb.trellis.gen;\
  1085. +        (*(m->cache_sad_tag))[(my)&7][(mx)&7] = pack16to32( (mx), (my) );\
  1086. +        (*(m->cache_sad_score))[(my)&7][(mx)&7] = cost;\
  1087. +    }\
  1088. +    cost += BITS_MVD(mx,my);\
  1089.      COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my );\
  1090.  }
  1091.  
  1092. @@ -123,6 +127,18 @@ static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_ite
  1093.      COPY3_IF_LT( bcost, costs[3], bmx, omx+(m3x), bmy, omy+(m3y) );\
  1094.  }
  1095.  
  1096. +#define CACHE_MV_X3( m0x, m0y, m1x, m1y, m2x, m2y ) {\
  1097. +    (*(m->cache_sad_gen))[(m0y)&7][(m0x)&7] = h->mb.trellis.gen;\
  1098. +    (*(m->cache_sad_gen))[(m1y)&7][(m1x)&7] = h->mb.trellis.gen;\
  1099. +    (*(m->cache_sad_gen))[(m2y)&7][(m2x)&7] = h->mb.trellis.gen;\
  1100. +    (*(m->cache_sad_tag))[(m0y)&7][(m0x)&7] = pack16to32( m0x, m0y );\
  1101. +    (*(m->cache_sad_tag))[(m1y)&7][(m1x)&7] = pack16to32( m1x, m1y );\
  1102. +    (*(m->cache_sad_tag))[(m2y)&7][(m2x)&7] = pack16to32( m2x, m2y );\
  1103. +    (*(m->cache_sad_score))[(m0y)&7][(m0x)&7] = costs[0];\
  1104. +    (*(m->cache_sad_score))[(m1y)&7][(m1x)&7] = costs[1];\
  1105. +    (*(m->cache_sad_score))[(m2y)&7][(m2x)&7] = costs[2];\
  1106. +}
  1107. +
  1108.  #define COST_MV_X3_ABS( m0x, m0y, m1x, m1y, m2x, m2y )\
  1109.  {\
  1110.      h->pixf.fpelcmp_x3[i_pixel]( p_fenc,\
  1111. @@ -130,6 +146,8 @@ static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_ite
  1112.          p_fref_w + (m1x) + (m1y)*stride,\
  1113.          p_fref_w + (m2x) + (m2y)*stride,\
  1114.          stride, costs );\
  1115. +    if ( h->param.analyse.b_cache_sads )\
  1116. +        CACHE_MV_X3( m0x, m0y, m1x, m1y, m2x, m2y )\
  1117.      costs[0] += p_cost_mvx[(m0x)<<2]; /* no cost_mvy */\
  1118.      costs[1] += p_cost_mvx[(m1x)<<2];\
  1119.      costs[2] += p_cost_mvx[(m2x)<<2];\
  1120. @@ -201,6 +219,10 @@ void x264_me_search_ref( x264_t *h, x264_me_t *m, int16_t (*mvc)[2], int i_mvc,
  1121.  #define pack16to32_mask2(mx,my) ((mx<<16)|(my&0x7FFF))
  1122.      uint32_t mv_min = pack16to32_mask2( -mv_x_min, -mv_y_min );
  1123.      uint32_t mv_max = pack16to32_mask2( mv_x_max, mv_y_max )|0x8000;
  1124. +    m->mv_min[0] = mv_x_min;
  1125. +    m->mv_min[1] = mv_y_min;
  1126. +    m->mv_max[0] = mv_x_max;
  1127. +    m->mv_max[1] = mv_y_max;
  1128.  
  1129.  #define CHECK_MVRANGE(mx,my) (!(((pack16to32_mask2(mx,my) + mv_min) | (mv_max - pack16to32_mask2(mx,my))) & 0x80004000))
  1130.  
  1131. diff --git a/encoder/me.h b/encoder/me.h
  1132. old mode 100644
  1133. new mode 100755
  1134. index 3281630..a843d67
  1135. --- a/encoder/me.h
  1136. +++ b/encoder/me.h
  1137. @@ -53,6 +53,12 @@ typedef struct
  1138.      int cost_mv;        /* lambda * nbits for the chosen mv */
  1139.      int cost;           /* satd + lambda * nbits */
  1140.      ALIGNED_4( int16_t mv[2] );
  1141. +    ALIGNED_4( int16_t mv_min[2] );
  1142. +    ALIGNED_4( int16_t mv_max[2] );
  1143. +    
  1144. +    uint32_t (*cache_sad_gen)[8][8];
  1145. +    uint32_t (*cache_sad_tag)[8][8];
  1146. +    int (*cache_sad_score)[8][8];
  1147.  } ALIGNED_16( x264_me_t );
  1148.  
  1149.  typedef struct
  1150. diff --git a/encoder/ratecontrol.c b/encoder/ratecontrol.c
  1151. old mode 100644
  1152. new mode 100755
  1153. diff --git a/encoder/ratecontrol.h b/encoder/ratecontrol.h
  1154. old mode 100644
  1155. new mode 100755
  1156. diff --git a/encoder/rdo.c b/encoder/rdo.c
  1157. old mode 100644
  1158. new mode 100755
  1159. diff --git a/encoder/set.c b/encoder/set.c
  1160. old mode 100644
  1161. new mode 100755
  1162. diff --git a/encoder/set.h b/encoder/set.h
  1163. old mode 100644
  1164. new mode 100755
  1165. diff --git a/encoder/slicetype.c b/encoder/slicetype.c
  1166. old mode 100644
  1167. new mode 100755
  1168. diff --git a/encoder/trellis.c b/encoder/trellis.c
  1169. new file mode 100755
  1170. index 0000000..4258bb5
  1171. --- /dev/null
  1172. +++ b/encoder/trellis.c
  1173. @@ -0,0 +1,1042 @@
  1174. +#include "common/common.h"
  1175. +
  1176. +// some constants for initializing structs
  1177. +const int16_t mv_zero[2] = {0,0};
  1178. +const mb_t mb_i_4x4 = {I_4x4, 0, 0, {0, 0}, {0, 0}, {0, 0, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1179. +const mb_t mb_i_8x8 = {I_8x8, 0, 0, {0, 0}, {0, 0}, {0, 0, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1180. +const mb_t mb_i_16x16 = {I_16x16, 0, 0, {0, 0}, {0, 0}, {0, 0, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1181. +const mb_t mb_p_16x16 = {P_L0, 0, 0, {0, 0}, {0, 0}, {D_16x16, 0, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1182. +const mb_t mb_p_16x8 = {P_L0, 0, 0, {0, 0}, {0, 0}, {D_16x8, D_16x8, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1183. +const mb_t mb_p_8x16 = {P_L0, 0, 0, {0, 0}, {0, 0}, {D_8x16, D_8x16, 0, 0}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1184. +const mb_t mb_p_8x8 = {P_8x8, 0, 0, {0, 0}, {0, 0}, {D_8x8, D_8x8, D_8x8, D_8x8}, {{{0,0}, {0,0}, {0,0}, {0,0}}, {{0,0}, {0,0}, {0,0}, {0,0}}}, {{0,0}, {0,0}, {0,0}, {0,0}}};
  1185. +
  1186. +int get_cached_fpel_sad( x264_t *h, int class, int partition, int i_list, int i_mv, int row, int col, int x, int y ) {
  1187. +    int result, valid, table, idx, idy;
  1188. +    uint32_t tag;
  1189. +    int i_mb = row * h->mb.i_mb_width + col;
  1190. +    uint32_t (*cache_gen)[8][8];
  1191. +    uint32_t (*cache_tag)[8][8];
  1192. +    int (*cache_score)[8][8];
  1193. +    x264_me_t *m;
  1194. +        
  1195. +    if ( IS_INTRA( class ) ) {
  1196. +        result = h->mb.trellis.cache_sad_intra[i_mb];
  1197. +    }
  1198. +    else {
  1199. +        valid = 1;
  1200. +        table = ( class == P_L0 && partition == D_16x16 ) ? ( 0 ) :
  1201. +                ( ( class == P_8x8 ) ? ( 1 + i_mv ) :
  1202. +                ( ( class == P_L0 && partition == D_16x8 ) ? ( 5 + i_mv ) :
  1203. +                ( 7 + i_mv ) ) );
  1204. +        tag = pack16to32( x, y );
  1205. +        cache_gen = &(h->mb.trellis.cache_sad_gen[i_list][i_mb][table]);
  1206. +        cache_tag = &(h->mb.trellis.cache_sad_tag[i_list][i_mb][table]);
  1207. +        cache_score = &(h->mb.trellis.cache_sad_score[i_list][i_mb][table]);
  1208. +        m = &(h->mb.trellis.cache_me_t[i_list][i_mb][table]);
  1209. +        const int i_pixel = m->i_pixel;
  1210. +        const int i_fref_stride = m->i_stride[0];
  1211. +        const int i_fenc_stride = h->fdec->i_stride[0];
  1212. +        const int xoff = ( class == P_L0 && partition == D_16x16 ) ? ( 0 ) :
  1213. +                ( ( class == P_8x8 ) ? ( 8*(i_mv&1) ) :
  1214. +                ( ( class == P_L0 && partition == D_16x8 ) ? ( 0 ) :
  1215. +                ( 8*i_mv ) ) );
  1216. +        const int yoff = ( class == P_L0 && partition == D_16x16 ) ? ( 0 ) :
  1217. +                ( ( class == P_8x8 ) ? ( 8*(i_mv>>1) ) :
  1218. +                ( ( class == P_L0 && partition == D_16x8 ) ? ( 8*i_mv ) :
  1219. +                ( 0 ) ) );
  1220. +        pixel *p_fenc = &(h->fenc->plane[0][16 * col + xoff + ( 16 * row + yoff) * i_fenc_stride]);
  1221. +        pixel *p_fref_w = m->p_fref_w;
  1222. +        idx = 7 & x;
  1223. +        idy = 7 & y;
  1224. +        
  1225. +        if ( (*cache_gen)[idy][idx] != h->mb.trellis.gen ) {
  1226. +            valid = 0;
  1227. +            (*cache_gen)[idy][idx] = h->mb.trellis.gen;
  1228. +        }
  1229. +        if ( (*cache_tag)[idy][idx] != tag ){
  1230. +            valid = 0;
  1231. +            (*cache_tag)[idy][idx] = tag;
  1232. +        }
  1233. +        
  1234. +        if ( ! valid ) {
  1235. +            (*cache_score)[idy][idx] = h->pixf.fpelcmp[i_pixel]( p_fenc, i_fenc_stride, &p_fref_w[(y)*i_fref_stride+(x)], i_fref_stride );
  1236. +        }
  1237. +            
  1238. +        result = (*cache_score)[idy][idx];
  1239. +    }
  1240. +
  1241. +    return result;
  1242. +}
  1243. +
  1244. +void collect_candidates( x264_t *h ) {
  1245. +    int row, col, boff;
  1246. +    int i_mb = 0;
  1247. +    int type, part;
  1248. +    int mb_stride = h->mb.trellis.mb_stride_col;
  1249. +    for ( row = 0; row < h->mb.i_mb_height; row++ ) {
  1250. +        for ( col = 0; col < h->mb.i_mb_width; col++ ) {
  1251. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x16]), &mb_p_16x16 );
  1252. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8]), &mb_p_8x8 );
  1253. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8]), &mb_p_16x8 );
  1254. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16]), &mb_p_8x16 );
  1255. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_INTRA]), ( h->mb.trellis.cache_mode_intra[i_mb] == I_16x16 ) ? ( &mb_i_16x16 ) : ( ( h->mb.trellis.cache_mode_intra[i_mb] == I_8x8 ) ? ( &mb_i_8x8 ) : ( &mb_i_4x4 ) ) );
  1256. +            
  1257. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x16].mv[0][0][0] = h->mb.trellis.cache_me_t[0][i_mb][0].mv[0] >> 2;
  1258. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x16].mv[0][0][1] = h->mb.trellis.cache_me_t[0][i_mb][0].mv[1] >> 2;
  1259. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][0][0] = h->mb.trellis.cache_me_t[0][i_mb][1].mv[0] >> 2;
  1260. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][0][1] = h->mb.trellis.cache_me_t[0][i_mb][1].mv[1] >> 2;
  1261. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][1][0] = h->mb.trellis.cache_me_t[0][i_mb][2].mv[0] >> 2;
  1262. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][1][1] = h->mb.trellis.cache_me_t[0][i_mb][2].mv[1] >> 2;
  1263. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][2][0] = h->mb.trellis.cache_me_t[0][i_mb][3].mv[0] >> 2;
  1264. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][2][1] = h->mb.trellis.cache_me_t[0][i_mb][3].mv[1] >> 2;
  1265. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][3][0] = h->mb.trellis.cache_me_t[0][i_mb][4].mv[0] >> 2;
  1266. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv[0][3][1] = h->mb.trellis.cache_me_t[0][i_mb][4].mv[1] >> 2;
  1267. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv[0][0][0] = h->mb.trellis.cache_me_t[0][i_mb][5].mv[0] >> 2;
  1268. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv[0][0][1] = h->mb.trellis.cache_me_t[0][i_mb][5].mv[1] >> 2;
  1269. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv[0][1][0] = h->mb.trellis.cache_me_t[0][i_mb][6].mv[0] >> 2;
  1270. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv[0][1][1] = h->mb.trellis.cache_me_t[0][i_mb][6].mv[1] >> 2;
  1271. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv[0][0][0] = h->mb.trellis.cache_me_t[0][i_mb][7].mv[0] >> 2;
  1272. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv[0][0][1] = h->mb.trellis.cache_me_t[0][i_mb][7].mv[1] >> 2;
  1273. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv[0][1][0] = h->mb.trellis.cache_me_t[0][i_mb][8].mv[0] >> 2;
  1274. +            h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv[0][1][1] = h->mb.trellis.cache_me_t[0][i_mb][8].mv[1] >> 2;
  1275. +
  1276. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x16].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][0].mv_min[0]) );
  1277. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][1].mv_min[0]) );
  1278. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][2].mv_min[0]) );
  1279. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][3].mv_min[0]) );
  1280. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][4].mv_min[0]) );
  1281. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][5].mv_min[0]) );
  1282. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][6].mv_min[0]) );
  1283. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][7].mv_min[0]) );
  1284. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv_min[0]), &(h->mb.trellis.cache_me_t[0][i_mb][8].mv_min[0]) );
  1285. +
  1286. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x16].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][0].mv_max[0]) );
  1287. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][1].mv_max[0]) );
  1288. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][2].mv_max[0]) );
  1289. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][3].mv_max[0]) );
  1290. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][4].mv_max[0]) );
  1291. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][5].mv_max[0]) );
  1292. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_16x8].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][6].mv_max[0]) );
  1293. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][7].mv_max[0]) );
  1294. +            CP_MV( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_8x16].mv_max[0]), &(h->mb.trellis.cache_me_t[0][i_mb][8].mv_max[0]) );
  1295. +            
  1296. +            type = h->mb.trellis.best_type[0][i_mb];
  1297. +            part = h->mb.trellis.best_part[0][i_mb];
  1298. +            boff = ( type == P_L0 && part == D_16x16 ) ? ( T_P_16x16 ) :
  1299. +                    ( ( type == P_8x8 && part == D_8x8 ) ? ( T_P_8x8 ) :
  1300. +                    ( ( type == P_L0 && part == D_16x8 ) ? ( T_P_16x8 ) :
  1301. +                    ( ( type == P_L0 && part == D_8x16 ) ? ( T_P_8x16 ) :
  1302. +                    ( T_P_INTRA ) ) ) );
  1303. +            CP_MB( &(h->mb.trellis.mb[i_mb*mb_stride+T_P_BEST]), &(h->mb.trellis.mb[i_mb*mb_stride+boff]) );
  1304. +
  1305. +            i_mb++;
  1306. +        }
  1307. +    }
  1308. +}
  1309. +
  1310. +void fill_me_trellis_params( x264_t *h, int mode, int direction ) {
  1311. +    h->mb.trellis.mode = mode;
  1312. +    h->mb.trellis.direction = direction;
  1313. +    h->mb.trellis.breadth = ( mode == MODE_COMBINED_DIAMOND ) ? COMBINED_DIAMOND_CANDIDATES :
  1314. +            ( ( mode == MODE_DECIDE ) ? DECIDE_CANDIDATES :
  1315. +            ( ( MODE_DIAMOND ) ? DIAMOND_CANDIDATES :
  1316. +            SQUARE2_CANDIDATES ) );
  1317. +    h->mb.trellis.length = ( direction == ROW_TRELLIS ) ? h->mb.i_mb_width : h->mb.i_mb_height;
  1318. +}
  1319. +
  1320. +void get_left_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv ) {
  1321. +    switch ( block->part[i_mv] ) {
  1322. +        case D_8x8 :
  1323. +            if ( i_mv == 0 )
  1324. +                CP_MV( pred, block->mv[i_list][1] );
  1325. +            else
  1326. +                CP_MV( pred, block->mv[i_list][3] );
  1327. +            break;
  1328. +        case D_16x8 :
  1329. +            if ( i_mv == 0 )
  1330. +                CP_MV( pred, block->mv[i_list][0] );
  1331. +            else
  1332. +                CP_MV( pred, block->mv[i_list][1] );
  1333. +            break;
  1334. +        case D_8x16 :
  1335. +            CP_MV( pred, block->mv[i_list][1] );
  1336. +            break;
  1337. +        default :
  1338. +            CP_MV( pred, block->mv[i_list][0] );
  1339. +            break;
  1340. +    }
  1341. +}
  1342. +
  1343. +void get_top_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv ) {
  1344. +    switch ( block->part[i_mv] ) {
  1345. +        case D_8x8 :
  1346. +            if ( i_mv == 0 )
  1347. +                CP_MV( pred, block->mv[i_list][2] );
  1348. +            else
  1349. +                CP_MV( pred, block->mv[i_list][3] );
  1350. +            break;
  1351. +        case D_16x8 :
  1352. +            CP_MV( pred, block->mv[i_list][1] );
  1353. +            break;
  1354. +        case D_8x16 :
  1355. +            if ( i_mv == 0 )
  1356. +                CP_MV( pred, block->mv[i_list][0] );
  1357. +            else
  1358. +                CP_MV( pred, block->mv[i_list][1] );
  1359. +            break;
  1360. +        default :
  1361. +            CP_MV( pred, block->mv[i_list][0] );
  1362. +            break;
  1363. +    }
  1364. +}
  1365. +
  1366. +void get_top_right_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv ) {
  1367. +    switch ( block->part[i_mv] ) {
  1368. +        case D_8x8 :
  1369. +            if ( i_mv == 0 )
  1370. +                CP_MV( pred, block->mv[i_list][3] );
  1371. +            else
  1372. +                CP_MV( pred, block->mv[i_list][2] );
  1373. +            break;
  1374. +        case D_16x8 :
  1375. +            CP_MV( pred, block->mv[i_list][1] );
  1376. +            break;
  1377. +        case D_8x16 :
  1378. +            if ( i_mv == 0 )
  1379. +                CP_MV( pred, block->mv[i_list][1] );
  1380. +            else
  1381. +                CP_MV( pred, block->mv[i_list][0] );
  1382. +            break;
  1383. +        default :
  1384. +            CP_MV( pred, block->mv[i_list][0] );
  1385. +            break;
  1386. +    }
  1387. +}
  1388. +
  1389. +void get_top_left_predictor( int16_t pred[2], mb_t *block, int i_list, int i_mv ) {
  1390. +    switch ( block->part[i_mv] ) {
  1391. +        case D_8x8 :
  1392. +            if ( i_mv == 0 )
  1393. +                CP_MV( pred, block->mv[i_list][3] );
  1394. +            else
  1395. +                CP_MV( pred, block->mv[i_list][2] );
  1396. +            break;
  1397. +        case D_16x8 :
  1398. +            CP_MV( pred, block->mv[i_list][1] );
  1399. +            break;
  1400. +        case D_8x16 :
  1401. +            if ( i_mv == 0 )
  1402. +                CP_MV( pred, block->mv[i_list][1] );
  1403. +            else
  1404. +                CP_MV( pred, block->mv[i_list][0] );
  1405. +            break;
  1406. +        default :
  1407. +            CP_MV( pred, block->mv[i_list][0] );
  1408. +            break;
  1409. +    }
  1410. +}
  1411. +
  1412. +void choose_predictors( x264_t *h, mb_t *pred, int row, int col, mb_t *current, mb_t *left, mb_t *top, mb_t *diag ) {
  1413. +    int16_t a[2], b[2], c[2], fetch[2];
  1414. +    int i_list;
  1415. +
  1416. +    switch ( current->class ) {
  1417. +        
  1418. +        case I_4x4 :
  1419. +        
  1420. +            CP_MB( pred, &mb_i_4x4 );
  1421. +            break;
  1422. +
  1423. +        case I_8x8 :
  1424. +        
  1425. +            CP_MB( pred, &mb_i_8x8 );
  1426. +            break;
  1427. +
  1428. +        case I_16x16 :
  1429. +        
  1430. +            CP_MB( pred, &mb_i_16x16 );
  1431. +            break;
  1432. +
  1433. +        case P_L0 :
  1434. +        
  1435. +            i_list = 0;
  1436. +            
  1437. +            switch( current->part[0] ) {
  1438. +
  1439. +                case D_16x16 :
  1440. +
  1441. +                    CP_MB( pred, &mb_p_16x16 );
  1442. +
  1443. +                    if ( col == 0 ) {
  1444. +                        CP_MV( a, mv_zero );
  1445. +                    }
  1446. +                    else {
  1447. +                        get_left_predictor( fetch, left, i_list, 0 );
  1448. +                        CP_MV( a, fetch );
  1449. +                    }
  1450. +
  1451. +                    if ( row == 0 ) {
  1452. +                        CP_MV( b, mv_zero );
  1453. +                        CP_MV( c, mv_zero );
  1454. +                    } else {
  1455. +                        get_top_predictor( fetch, top, i_list, 0 );
  1456. +                        CP_MV( b, fetch );
  1457. +
  1458. +                        if ( col == h->mb.i_mb_width - 1 )
  1459. +                            get_top_left_predictor( fetch, diag, i_list, 0 );
  1460. +                        else
  1461. +                            get_top_right_predictor( fetch, diag, i_list, 1 );
  1462. +                        CP_MV( c, fetch );
  1463. +                    }
  1464. +
  1465. +                    pred->mv[i_list][0][0] = X264_MED( a[0], b[0], c[0] );
  1466. +                    pred->mv[i_list][0][1] = X264_MED( a[1], b[1], c[1] );
  1467. +
  1468. +                    break;
  1469. +
  1470. +                case D_16x8 :
  1471. +
  1472. +                    CP_MB( pred, &mb_p_16x8 );
  1473. +
  1474. +                    if ( row == 0 )
  1475. +                        CP_MV( pred->mv[i_list][0], mv_zero );
  1476. +                    else {
  1477. +                        get_top_predictor( fetch, top, i_list, 0 );
  1478. +                        CP_MV( pred->mv[i_list][0], fetch );
  1479. +                    }
  1480. +
  1481. +                    if ( col == 0 )
  1482. +                        CP_MV( pred->mv[i_list][1], mv_zero );
  1483. +                    else {
  1484. +                        get_left_predictor( fetch, left, i_list, 1 );
  1485. +                        CP_MV( pred->mv[i_list][1], fetch );
  1486. +                    }
  1487. +
  1488. +                    break;
  1489. +                    
  1490. +                case D_8x16 :
  1491. +
  1492. +                    CP_MB( pred, &mb_p_8x16 );
  1493. +
  1494. +                    if ( col == 0 )
  1495. +                        CP_MV( pred->mv[i_list][0], mv_zero );
  1496. +                    else {
  1497. +                        get_left_predictor( fetch, left, i_list, 0 );
  1498. +                        CP_MV( pred->mv[i_list][0], fetch );
  1499. +                    }
  1500. +
  1501. +                    if ( row == 0 )
  1502. +                        CP_MV( pred->mv[i_list][1], mv_zero );
  1503. +                    else {
  1504. +                        if ( col == h->mb.i_mb_width-1 )
  1505. +                            get_top_left_predictor( fetch, top, i_list, 1 );
  1506. +                        else
  1507. +                            get_top_right_predictor( fetch, diag, i_list, 1 );
  1508. +                        CP_MV( pred->mv[i_list][1], fetch );
  1509. +                    }
  1510. +
  1511. +                    break;
  1512. +                    
  1513. +            }
  1514. +
  1515. +        case P_8x8 :
  1516. +        
  1517. +            i_list = 0;
  1518. +            CP_MB( pred, &mb_p_8x8 );
  1519. +
  1520. +            for ( int i = 0; i < 4; i++ ) {
  1521. +
  1522. +                if ( col == 0 && !( i & 1 ) )
  1523. +                    CP_MV( a, mv_zero );
  1524. +                else {
  1525. +                    if ( i & 1 ) {
  1526. +                        if ( i == 1 )
  1527. +                            CP_MV( fetch, current->mv[i_list][0] );
  1528. +                        else
  1529. +                            CP_MV( fetch, current->mv[i_list][2] );
  1530. +                    }
  1531. +                    else
  1532. +                        get_left_predictor( fetch, left, i_list, i );
  1533. +                        
  1534. +                    CP_MV( a, fetch );
  1535. +                }
  1536. +
  1537. +                if ( row == 0 && i < 2) {
  1538. +                    CP_MV( b, mv_zero );
  1539. +                    CP_MV( c, mv_zero );
  1540. +                } else {
  1541. +                    if ( i > 1 ) {
  1542. +                        if ( i == 2 )
  1543. +                            CP_MV( fetch, current->mv[i_list][0] );
  1544. +                        else
  1545. +                            CP_MV( fetch, current->mv[i_list][1] );
  1546. +                    }
  1547. +                    else
  1548. +                        get_top_predictor( fetch, top, i_list, i );
  1549. +                    CP_MV( b, fetch );
  1550. +
  1551. +                    
  1552. +                    if ( i == 0 )
  1553. +                        get_top_right_predictor( fetch, top, i_list, i );
  1554. +                    else if ( i == 1 ) {
  1555. +                        if ( col == h->mb.i_mb_width - 1 )
  1556. +                            get_top_left_predictor( fetch, diag, i_list, i );
  1557. +                        else
  1558. +                            get_top_right_predictor( fetch, diag, i_list, i );
  1559. +                    }
  1560. +                    else if ( i == 2 )
  1561. +                        CP_MV( fetch, current->mv[i_list][1] );
  1562. +                    else
  1563. +                        CP_MV( fetch, current->mv[i_list][0] );
  1564. +
  1565. +                    CP_MV( c, fetch );
  1566. +                }
  1567. +
  1568. +                pred->mv[i_list][i][0] = X264_MED( a[0], b[0], c[0] );
  1569. +                pred->mv[i_list][i][1] = X264_MED( a[1], b[1], c[1] );
  1570. +
  1571. +            }
  1572. +
  1573. +            break;
  1574. +
  1575. +    }
  1576. +
  1577. +}
  1578. +
  1579. +int cost_mb( x264_t *h, int row, int col, int i_mv_max, mb_t *block, mb_t *left, mb_t *top, mb_t *diag ) {
  1580. +    mb_t pred;
  1581. +    choose_predictors( h, &pred, row, col, block, left, top, diag );
  1582. +    block->costSAD = 0;
  1583. +    block->costMV = 0;
  1584. +    int i, x, y, px, py, i_list;
  1585. +    
  1586. +
  1587. +    switch ( block->class ) {
  1588. +
  1589. +        case I_4x4 :
  1590. +        case I_8x8 :
  1591. +        case I_16x16 :
  1592. +        case I_PCM :
  1593. +        
  1594. +        
  1595. +            block->costSAD += get_cached_fpel_sad( h, block->class, 0, 0, 0, row, col, 0, 0 );
  1596. +            break;
  1597. +            
  1598. +        case P_L0 :
  1599. +        
  1600. +            i_list = 0;
  1601. +        
  1602. +            switch ( block->part[0] ) {
  1603. +
  1604. +                case D_16x16 :
  1605. +
  1606. +                    x = block->mv[i_list][0][0];
  1607. +                    y = block->mv[i_list][0][1];
  1608. +                    px = pred.mv[i_list][0][0];
  1609. +                    py = pred.mv[i_list][0][1];
  1610. +                    block->costSAD += get_cached_fpel_sad( h, P_L0, D_16x16, 0, 0, row, col, x, y );
  1611. +                    block->costMV += T_BITS_MVD( x-px, y-py );
  1612. +                    break;
  1613. +
  1614. +                case D_16x8 :
  1615. +
  1616. +                    for ( i = 0; i < X264_MIN(i_mv_max+1,2); i++ ) {
  1617. +                        x = block->mv[i_list][i][0];
  1618. +                        y = block->mv[i_list][i][1];
  1619. +                        px = pred.mv[i_list][i][0];
  1620. +                        py = pred.mv[i_list][i][1];
  1621. +                        block->costSAD += get_cached_fpel_sad( h, P_L0, D_16x8, 0, i, row, col, x, y );
  1622. +                        block->costMV += T_BITS_MVD( x-px, y-py );
  1623. +                    }
  1624. +                    break;
  1625. +
  1626. +                case D_8x16 :
  1627. +
  1628. +                    for ( i = 0; i < X264_MIN(i_mv_max+1,2); i++ ) {
  1629. +                        x = block->mv[i_list][i][0];
  1630. +                        y = block->mv[i_list][i][1];
  1631. +                        px = pred.mv[i_list][i][0];
  1632. +                        py = pred.mv[i_list][i][1];
  1633. +                        block->costSAD += get_cached_fpel_sad( h, P_L0, D_8x16, 0, i, row, col, x, y );
  1634. +                        block->costMV += T_BITS_MVD( x-px, y-py );
  1635. +                    }
  1636. +                    break;
  1637. +                    
  1638. +            }
  1639. +            
  1640. +            break;
  1641. +
  1642. +        case P_8x8 :
  1643. +        
  1644. +            i_list = 0;
  1645. +
  1646. +            for ( i = 0; i < X264_MIN(i_mv_max+1,4); i++ ) {
  1647. +                x = block->mv[i_list][i][0];
  1648. +                y = block->mv[i_list][i][1];
  1649. +                px = pred.mv[i_list][i][0];
  1650. +                py = pred.mv[i_list][i][1];
  1651. +                block->costSAD += get_cached_fpel_sad( h, P_8x8, D_8x8, 0, i, row, col, x, y );
  1652. +                block->costMV += T_BITS_MVD( x-px, y-py );
  1653. +            }
  1654. +            break;
  1655. +
  1656. +    }
  1657. +
  1658. +    //block->costMV += T_BITS_MODE( block->type );
  1659. +    
  1660. +    return ( block->costSAD + block->costMV );
  1661. +}
  1662. +
  1663. +int sad_cost( x264_t *h ) {
  1664. +    int row, col;
  1665. +    int cost;
  1666. +    int i_list, offCC, offCL, offTL, offTR, offTC;
  1667. +    int mb_stride_col = h->mb.trellis.mb_stride_col;
  1668. +    int mb_stride_row = h->mb.trellis.mb_stride_row;
  1669. +    mb_t *block, *left, *top, *diag;
  1670. +
  1671. +    cost = 0;
  1672. +    i_list = 0;
  1673. +    offCC = 0;
  1674. +    offCL = offCC - mb_stride_col;
  1675. +    offTC = offCC - mb_stride_row;
  1676. +    offTL = offTC - mb_stride_col;
  1677. +    offTR = offTC + mb_stride_col;
  1678. +    for ( row = 0; row < h->mb.i_mb_height; row++ ) {
  1679. +        for ( col = 0; col < h->mb.i_mb_width; col++ ) {
  1680. +            block = &(h->mb.trellis.mb[offCC+T_P_BEST]);
  1681. +            left = ( col == 0 ) ? NULL : &(h->mb.trellis.mb[offCL+T_P_BEST]);
  1682. +            top = ( row == 0 ) ? NULL : &(h->mb.trellis.mb[offTC+T_P_BEST]);
  1683. +            diag = ( row == 0 ) ? NULL :
  1684. +                    ( col == h->mb.i_mb_width - 1 ) ? &(h->mb.trellis.mb[offTL+T_P_BEST]) :
  1685. +                    &(h->mb.trellis.mb[offTR+T_P_BEST]);
  1686. +
  1687. +            cost_mb( h, row, col, 3, block, left, top, diag );
  1688. +            cost += block->costSAD;
  1689. +
  1690. +            offCC += mb_stride_col;
  1691. +            offCL += mb_stride_col;
  1692. +            offTC += mb_stride_col;
  1693. +            offTL += mb_stride_col;
  1694. +            offTR += mb_stride_col;
  1695. +        }
  1696. +    }
  1697. +
  1698. +    return cost;
  1699. +}
  1700. +
  1701. +int mv_cost( x264_t *h ) {
  1702. +    int row, col;
  1703. +    int cost;
  1704. +    int i_list, offCC, offCL, offTL, offTR, offTC;
  1705. +    int mb_stride_col = h->mb.trellis.mb_stride_col;
  1706. +    int mb_stride_row = h->mb.trellis.mb_stride_row;
  1707. +    mb_t *block, *left, *top, *diag;
  1708. +
  1709. +    cost = 0;
  1710. +    i_list = 0;
  1711. +    offCC = 0;
  1712. +    offCL = offCC - mb_stride_col;
  1713. +    offTC = offCC - mb_stride_row;
  1714. +    offTL = offTC - mb_stride_col;
  1715. +    offTR = offTC + mb_stride_col;
  1716. +    for ( row = 0; row < h->mb.i_mb_height; row++ ) {
  1717. +        for ( col = 0; col < h->mb.i_mb_width; col++ ) {
  1718. +            block = &(h->mb.trellis.mb[offCC+T_P_BEST]);
  1719. +            left = ( col == 0 ) ? NULL : &(h->mb.trellis.mb[offCL+T_P_BEST]);
  1720. +            top = ( row == 0 ) ? NULL : &(h->mb.trellis.mb[offTC+T_P_BEST]);
  1721. +            diag = ( row == 0 ) ? NULL :
  1722. +                    ( col == h->mb.i_mb_width - 1 ) ? &(h->mb.trellis.mb[offTL+T_P_BEST]) :
  1723. +                    &(h->mb.trellis.mb[offTR+T_P_BEST]);
  1724. +
  1725. +            cost_mb( h, row, col, 3, block, left, top, diag );
  1726. +            cost += block->costMV;
  1727. +
  1728. +            offCC += mb_stride_col;
  1729. +            offCL += mb_stride_col;
  1730. +            offTC += mb_stride_col;
  1731. +            offTL += mb_stride_col;
  1732. +            offTR += mb_stride_col;
  1733. +        }
  1734. +    }
  1735. +
  1736. +    return cost;
  1737. +}
  1738. +
  1739. +void fill_trellis( x264_t *h, int i_frame ) {
  1740. +    int i_trellis, i, j;
  1741. +    int mb_stride_col = h->mb.trellis.mb_stride_col;
  1742. +    int mb_stride_row = h->mb.trellis.mb_stride_row;
  1743. +    int t_stride = h->mb.trellis.t_stride;
  1744. +    int offR = ( h->mb.trellis.direction == ROW_TRELLIS ) ? i_frame * mb_stride_row : 0 ;
  1745. +    int offC = ( h->mb.trellis.direction == ROW_TRELLIS ) ? 0 : i_frame * mb_stride_col ;
  1746. +
  1747. +    if ( h->mb.trellis.mode == MODE_COMBINED_DIAMOND ) {
  1748. +        for ( i_trellis = 0; i_trellis < h->mb.trellis.length; i_trellis++ ) {
  1749. +            for ( i = 0; i < T_P_INTRA; i++ ) {
  1750. +                for ( j = 0; j < DIAMOND_CANDIDATES; j++ ) {
  1751. +                    CP_MB( &(h->mb.trellis.t[i_trellis*t_stride+i*DIAMOND_CANDIDATES+j].block), &(h->mb.trellis.mb[offR+offC+i]) );
  1752. +                }
  1753. +            }
  1754. +            CP_MB( &(h->mb.trellis.t[i_trellis*t_stride+T_P_INTRA*DIAMOND_CANDIDATES].block), &(h->mb.trellis.mb[offR+offC+T_P_INTRA]) );
  1755. +            
  1756. +            if ( h->mb.trellis.direction == ROW_TRELLIS )
  1757. +                offC += mb_stride_col;
  1758. +            else
  1759. +                offR += mb_stride_row;
  1760. +        }
  1761. +    }
  1762. +    else {
  1763. +        for ( i_trellis = 0; i_trellis < h->mb.trellis.length; i_trellis++ ) {
  1764. +            for ( j = 0; j < h->mb.trellis.breadth; j++ ) {
  1765. +                if ( ( h->mb.trellis.mode == MODE_DIAMOND ) || ( h->mb.trellis.mode == MODE_SQUARE2 ) )
  1766. +                    CP_MB( &(h->mb.trellis.t[i_trellis*t_stride+j].block), &(h->mb.trellis.mb[offR+offC+T_P_BEST]) );
  1767. +                else
  1768. +                    CP_MB( &(h->mb.trellis.t[i_trellis*t_stride+j].block), &(h->mb.trellis.mb[offR+offC+j]) );
  1769. +            }
  1770. +
  1771. +            if ( h->mb.trellis.direction == ROW_TRELLIS )
  1772. +                offC += mb_stride_col;
  1773. +            else
  1774. +                offR += mb_stride_row;
  1775. +        }
  1776. +    }
  1777. +}
  1778. +
  1779. +void adjust_trellis( x264_t *h, int sub ) {
  1780. +    mb_t *block;
  1781. +    int i_trellis, i_mv, partition, i_list;
  1782. +    int t_stride = h->mb.trellis.t_stride;
  1783. +    int length = (h->mb.trellis.direction == ROW_TRELLIS) ? h->mb.i_mb_width : h->mb.i_mb_height;
  1784. +    int i8x8 = (h->mb.trellis.direction == ROW_TRELLIS) ? sub*2 : sub;
  1785. +    int i16x8 = (h->mb.trellis.direction == ROW_TRELLIS) ? sub : 0;
  1786. +    int i8x16 = (h->mb.trellis.direction == ROW_TRELLIS) ? 0 : sub;
  1787. +
  1788. +    i_list = 0;
  1789. +    if ( h->mb.trellis.mode == MODE_COMBINED_DIAMOND ) {
  1790. +        for ( i_trellis = 0; i_trellis < length; i_trellis++ ) {
  1791. +            // bottom
  1792. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x16*DIAMOND_CANDIDATES+B___].block);
  1793. +            block->mv[i_list][0][1] = X264_MIN( block->mv[i_list][0][1] + 1, block->mv_max[1] );
  1794. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x8*DIAMOND_CANDIDATES+B___].block);
  1795. +            block->mv[i_list][i8x8][1] = X264_MIN( block->mv[i_list][i8x8][1] + 1, block->mv_max[1] );
  1796. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x8*DIAMOND_CANDIDATES+B___].block);
  1797. +            block->mv[i_list][i16x8][1] = X264_MIN( block->mv[i_list][i16x8][1] + 1, block->mv_max[1] );
  1798. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x16*DIAMOND_CANDIDATES+B___].block);
  1799. +            block->mv[i_list][i8x16][1] = X264_MIN( block->mv[i_list][i8x16][1] + 1, block->mv_max[1] );
  1800. +
  1801. +            // top
  1802. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x16*DIAMOND_CANDIDATES+T___].block);
  1803. +            block->mv[i_list][0][1] = X264_MAX( block->mv[i_list][0][1] - 1, block->mv_min[1] );
  1804. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x8*DIAMOND_CANDIDATES+T___].block);
  1805. +            block->mv[i_list][i8x8][1] = X264_MAX( block->mv[i_list][i8x8][1] - 1, block->mv_min[1] );
  1806. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x8*DIAMOND_CANDIDATES+T___].block);
  1807. +            block->mv[i_list][i16x8][1] = X264_MAX( block->mv[i_list][i16x8][1] - 1, block->mv_min[1] );
  1808. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x16*DIAMOND_CANDIDATES+T___].block);
  1809. +            block->mv[i_list][i8x16][1] = X264_MAX( block->mv[i_list][i8x16][1] - 1, block->mv_min[1] );
  1810. +
  1811. +            // left
  1812. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x16*DIAMOND_CANDIDATES+L___].block);
  1813. +            block->mv[i_list][0][0] = X264_MAX( block->mv[i_list][0][0] - 1, block->mv_min[0] );
  1814. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x8*DIAMOND_CANDIDATES+L___].block);
  1815. +            block->mv[i_list][i8x8][0] = X264_MAX( block->mv[i_list][i8x8][0] - 1, block->mv_min[0] );
  1816. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x8*DIAMOND_CANDIDATES+L___].block);
  1817. +            block->mv[i_list][i16x8][0] = X264_MAX( block->mv[i_list][i16x8][0] - 1, block->mv_min[0] );
  1818. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x16*DIAMOND_CANDIDATES+L___].block);
  1819. +            block->mv[i_list][i8x16][0] = X264_MAX( block->mv[i_list][i8x16][0] - 1, block->mv_min[0] );
  1820. +
  1821. +            // right
  1822. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x16*DIAMOND_CANDIDATES+R___].block);
  1823. +            block->mv[i_list][0][0] = X264_MIN( block->mv[i_list][0][0] + 1, block->mv_max[0] );
  1824. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x8*DIAMOND_CANDIDATES+R___].block);
  1825. +            block->mv[i_list][i8x8][0] = X264_MIN( block->mv[i_list][i8x8][0] + 1, block->mv_max[0] );
  1826. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_16x8*DIAMOND_CANDIDATES+R___].block);
  1827. +            block->mv[i_list][i16x8][0] = X264_MIN( block->mv[i_list][i16x8][0] + 1, block->mv_max[0] );
  1828. +            block = &(h->mb.trellis.t[i_trellis*t_stride+T_P_8x16*DIAMOND_CANDIDATES+R___].block);
  1829. +            block->mv[i_list][i8x16][0] = X264_MIN( block->mv[i_list][i8x16][0] + 1, block->mv_max[0] );
  1830. +        }
  1831. +    }
  1832. +    else if ( h->mb.trellis.mode == MODE_SQUARE2 ) {
  1833. +        for ( i_trellis = 0; i_trellis < length; i_trellis++ ) {
  1834. +            if ( ! IS_INTRA( h->mb.trellis.t[i_trellis*t_stride+C___].block.class ) ) {
  1835. +                partition = h->mb.trellis.t[i_trellis*t_stride+C___].block.part[0];
  1836. +                i_mv = ( ( partition == D_16x16 ) ? 0 : ( ( partition == D_8x16 ) ? i8x16 : ( ( partition == D_8x8 ) ? i8x8 : i16x8 ) ) );
  1837. +                block = &(h->mb.trellis.t[i_trellis*t_stride+B___].block);
  1838. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1839. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BL__].block);
  1840. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1841. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BLL_].block);
  1842. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1843. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BR__].block);
  1844. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1845. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BRR_].block);
  1846. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1847. +                block = &(h->mb.trellis.t[i_trellis*t_stride+T___].block);
  1848. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1849. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TL__].block);
  1850. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1851. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TLL_].block);
  1852. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1853. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TR__].block);
  1854. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1855. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TRR_].block);
  1856. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1857. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BB__].block);
  1858. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 2, block->mv_max[1] );
  1859. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBL_].block);
  1860. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 2, block->mv_max[1] );
  1861. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBLL].block);
  1862. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 2, block->mv_max[1] );
  1863. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBR_].block);
  1864. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 2, block->mv_max[1] );
  1865. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBRR].block);
  1866. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 2, block->mv_max[1] );
  1867. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TT__].block);
  1868. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 2, block->mv_min[1] );
  1869. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTL_].block);
  1870. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 2, block->mv_min[1] );
  1871. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTLL].block);
  1872. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 2, block->mv_min[1] );
  1873. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTR_].block);
  1874. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 2, block->mv_min[1] );
  1875. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTRR].block);
  1876. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 2, block->mv_min[1] );
  1877. +                block = &(h->mb.trellis.t[i_trellis*t_stride+L___].block);
  1878. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1879. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TL__].block);
  1880. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1881. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTL_].block);
  1882. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1883. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BL__].block);
  1884. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1885. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBL_].block);
  1886. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1887. +                block = &(h->mb.trellis.t[i_trellis*t_stride+R___].block);
  1888. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1889. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TR__].block);
  1890. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1891. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTR_].block);
  1892. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1893. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BR__].block);
  1894. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1895. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBR_].block);
  1896. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1897. +                block = &(h->mb.trellis.t[i_trellis*t_stride+LL__].block);
  1898. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 2, block->mv_min[0] );
  1899. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TLL_].block);
  1900. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 2, block->mv_min[0] );
  1901. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTLL].block);
  1902. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 2, block->mv_min[0] );
  1903. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BLL_].block);
  1904. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 2, block->mv_min[0] );
  1905. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBLL].block);
  1906. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 2, block->mv_min[0] );
  1907. +                block = &(h->mb.trellis.t[i_trellis*t_stride+RR__].block);
  1908. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 2, block->mv_max[0] );
  1909. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TRR_].block);
  1910. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 2, block->mv_max[0] );
  1911. +                block = &(h->mb.trellis.t[i_trellis*t_stride+TTRR].block);
  1912. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 2, block->mv_max[0] );
  1913. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BRR_].block);
  1914. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 2, block->mv_max[0] );
  1915. +                block = &(h->mb.trellis.t[i_trellis*t_stride+BBRR].block);
  1916. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 2, block->mv_max[0] );
  1917. +            }
  1918. +        }
  1919. +    }
  1920. +    else {
  1921. +        for ( i_trellis = 0; i_trellis < length; i_trellis++ ) {
  1922. +            if ( ! IS_INTRA( h->mb.trellis.t[i_trellis*t_stride+C___].block.class ) ) {
  1923. +                partition = h->mb.trellis.t[i_trellis*t_stride+C___].block.part[0];
  1924. +                i_mv = ( ( partition == D_16x16 ) ? 0 : ( ( partition == D_8x16 ) ? i8x16 : ( ( partition == D_8x8 ) ? i8x8 : i16x8 ) ) );
  1925. +                block = &(h->mb.trellis.t[i_trellis*t_stride+B___].block);
  1926. +                block->mv[i_list][i_mv][1] = X264_MIN( block->mv[i_list][i_mv][1] + 1, block->mv_max[1] );
  1927. +                block = &(h->mb.trellis.t[i_trellis*t_stride+T___].block);
  1928. +                block->mv[i_list][i_mv][1] = X264_MAX( block->mv[i_list][i_mv][1] - 1, block->mv_min[1] );
  1929. +                block = &(h->mb.trellis.t[i_trellis*t_stride+L___].block);
  1930. +                block->mv[i_list][i_mv][0] = X264_MAX( block->mv[i_list][i_mv][0] - 1, block->mv_min[0] );
  1931. +                block = &(h->mb.trellis.t[i_trellis*t_stride+R___].block);
  1932. +                block->mv[i_list][i_mv][0] = X264_MIN( block->mv[i_list][i_mv][0] + 1, block->mv_max[0] );
  1933. +            }
  1934. +        }
  1935. +    }
  1936. +}
  1937. +
  1938. +void generate_alternate_mvs( x264_t *h, int16_t mv[SQUARE2_CANDIDATES][2], mb_t *block, int i_list, int i_mv ) {
  1939. +    int i;
  1940. +
  1941. +    if ( h->mb.trellis.mode == MODE_SQUARE2 ) {
  1942. +        for ( i = 0; i < SQUARE2_CANDIDATES; i++ )
  1943. +            CP_MV( mv[i], block->mv[i_list][i_mv] );
  1944. +
  1945. +        mv[B___][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1946. +        mv[BLL_][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1947. +        mv[BL__][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1948. +        mv[BR__][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1949. +        mv[BRR_][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1950. +        mv[T___][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1951. +        mv[TL__][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1952. +        mv[TLL_][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1953. +        mv[TR__][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1954. +        mv[TRR_][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1955. +        mv[BB__][1] = X264_MIN( mv[C___][1] + 2, block->mv_max[1] );
  1956. +        mv[BBL_][1] = X264_MIN( mv[C___][1] + 2, block->mv_max[1] );
  1957. +        mv[BBLL][1] = X264_MIN( mv[C___][1] + 2, block->mv_max[1] );
  1958. +        mv[BBR_][1] = X264_MIN( mv[C___][1] + 2, block->mv_max[1] );
  1959. +        mv[BBRR][1] = X264_MIN( mv[C___][1] + 2, block->mv_max[1] );
  1960. +        mv[TT__][1] = X264_MAX( mv[C___][1] - 2, block->mv_min[1] );
  1961. +        mv[TTL_][1] = X264_MAX( mv[C___][1] - 2, block->mv_min[1] );
  1962. +        mv[TTLL][1] = X264_MAX( mv[C___][1] - 2, block->mv_min[1] );
  1963. +        mv[TTR_][1] = X264_MAX( mv[C___][1] - 2, block->mv_min[1] );
  1964. +        mv[TTRR][1] = X264_MAX( mv[C___][1] - 2, block->mv_min[1] );
  1965. +        mv[L___][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1966. +        mv[TL__][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1967. +        mv[TTL_][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1968. +        mv[BL__][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1969. +        mv[BBL_][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1970. +        mv[R___][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1971. +        mv[TR__][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1972. +        mv[TTR_][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1973. +        mv[BR__][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1974. +        mv[BBR_][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1975. +        mv[LL__][0] = X264_MAX( mv[C___][0] - 2, block->mv_min[0] );
  1976. +        mv[TLL_][0] = X264_MAX( mv[C___][0] - 2, block->mv_min[0] );
  1977. +        mv[TTLL][0] = X264_MAX( mv[C___][0] - 2, block->mv_min[0] );
  1978. +        mv[BLL_][0] = X264_MAX( mv[C___][0] - 2, block->mv_min[0] );
  1979. +        mv[BBLL][0] = X264_MAX( mv[C___][0] - 2, block->mv_min[0] );
  1980. +        mv[RR__][0] = X264_MIN( mv[C___][0] + 2, block->mv_max[0] );
  1981. +        mv[TRR_][0] = X264_MIN( mv[C___][0] + 2, block->mv_max[0] );
  1982. +        mv[TTRR][0] = X264_MIN( mv[C___][0] + 2, block->mv_max[0] );
  1983. +        mv[BRR_][0] = X264_MIN( mv[C___][0] + 2, block->mv_max[0] );
  1984. +        mv[BBRR][0] = X264_MIN( mv[C___][0] + 2, block->mv_max[0] );
  1985. +    }
  1986. +    else {
  1987. +        for ( i = 0; i < DIAMOND_CANDIDATES; i++ )
  1988. +            CP_MV( mv[i], block->mv[i_list][i_mv] );
  1989. +
  1990. +        mv[B___][1] = X264_MIN( mv[C___][1] + 1, block->mv_max[1] );
  1991. +        mv[T___][1] = X264_MAX( mv[C___][1] - 1, block->mv_min[1] );
  1992. +        mv[L___][0] = X264_MAX( mv[C___][0] - 1, block->mv_min[0] );
  1993. +        mv[R___][0] = X264_MIN( mv[C___][0] + 1, block->mv_max[0] );
  1994. +    }
  1995. +}
  1996. +
  1997. +void run_row_trellis( x264_t *h, int iter ) {
  1998. +    int i, j, k, m, best, row, col, subrow, i_adj, class, part, alt_states;
  1999. +    int cost, min_cost;
  2000. +    int i_list, offCC, offCL, offTL, offTR, offTC, offDC, offDL, offDR, offCR;
  2001. +    int mb_stride_col = h->mb.trellis.mb_stride_col;
  2002. +    int mb_stride_row = h->mb.trellis.mb_stride_row;
  2003. +    int t_stride = h->mb.trellis.t_stride;
  2004. +    int frameCostMVOld, frameCostMVNew, frameCostSADOld, frameCostSADNew;
  2005. +    int16_t alt[SQUARE2_CANDIDATES][2];
  2006. +    int t_type;
  2007. +    
  2008. +    i_list = 0;
  2009. +
  2010. +    for ( i = 0; i < iter; i++ ) {
  2011. +        for ( row = 0; row < h->mb.i_mb_height; row ++ ) {
  2012. +            for ( subrow = 0; subrow <= 1; subrow++ ) {
  2013. +
  2014. +                offCC = row * mb_stride_row;
  2015. +                for ( col = 0; col < h->mb.i_mb_width; col++ ) {
  2016. +                    for ( j = 0; j <= T_P_BEST; j++ )
  2017. +                        CP_MB( &(h->mb.trellis.backup[col*mb_stride_col+j]), &(h->mb.trellis.mb[offCC+j]) );
  2018. +                    offCC += mb_stride_col;
  2019. +                }
  2020. +
  2021. +                frameCostMVOld = mv_cost( h );
  2022. +                frameCostSADOld = sad_cost( h );
  2023. +
  2024. +                fill_trellis( h, row );
  2025. +                if ( h->mb.trellis.mode != MODE_DECIDE )
  2026. +                    adjust_trellis( h, subrow );
  2027. +                                    
  2028. +                offCC = row * mb_stride_row ;
  2029. +                offCL = offCC - mb_stride_col;
  2030. +                offCR = offCC + mb_stride_col;
  2031. +                offTC = offCC - mb_stride_row;
  2032. +                offDC = offCC + mb_stride_row;
  2033. +                offTL = offTC - mb_stride_col;
  2034. +                offDL = offDC - mb_stride_col;
  2035. +                offTR = offTC + mb_stride_col;
  2036. +                offDR = offDC + mb_stride_col;
  2037. +                for ( j = 0; j < h->mb.trellis.breadth; j++ ) {
  2038. +                    h->mb.trellis.t[j].costCC = cost_mb( h, row, 0, 3, &(h->mb.trellis.t[j].block), NULL, ( row == 0 ) ? NULL : &(h->mb.trellis.mb[offTC+T_P_BEST]), ( row == 0 ) ? NULL : &(h->mb.trellis.mb[offTR+T_P_BEST]) );
  2039. +                    h->mb.trellis.t[j].costDC = ( row == h->mb.i_mb_height-1 ) ? 0 : cost_mb( h, row+1, 0, 3, &(h->mb.trellis.mb[offDC+T_P_BEST]), NULL, &(h->mb.trellis.t[j].block), &(h->mb.trellis.mb[offCR+T_P_BEST]) );
  2040. +                    h->mb.trellis.t[j].costDL = 0;
  2041. +                    h->mb.trellis.t[j].costDR = 0;
  2042. +                    h->mb.trellis.t[j].accCost = h->mb.trellis.t[j].costCC + h->mb.trellis.t[j].costDC;
  2043. +                }
  2044. +
  2045. +                offCC += mb_stride_col;
  2046. +                offCL += mb_stride_col;
  2047. +                offCR += mb_stride_col;
  2048. +                offTC += mb_stride_col;
  2049. +                offDC += mb_stride_col;
  2050. +                offTL += mb_stride_col;
  2051. +                offDL += mb_stride_col;
  2052. +                offTR += mb_stride_col;
  2053. +                offDR += mb_stride_col;
  2054. +                
  2055. +                for ( col = 1; col < h->mb.trellis.length; col++ ) {
  2056. +                    for ( j = 0; j < h->mb.trellis.breadth; j++ ) {
  2057. +                        min_cost = INT_MAX;
  2058. +                        class = h->mb.trellis.t[col*t_stride+j].block.class;
  2059. +                        part = h->mb.trellis.t[col*t_stride+j].block.part[0];
  2060. +                        i_adj = ( subrow && h->mb.trellis.t[col*t_stride+j].block.class == P_8x8 ) ? 3 : 1;
  2061. +                        alt_states = ( class == P_8x8 || ( class == P_L0 && part == D_8x16 ) ) ? ( ( h->mb.trellis.mode == MODE_SQUARE2 ) ? SQUARE2_CANDIDATES : DIAMOND_CANDIDATES ) : 1;
  2062. +                        if ( class == P_8x8 || ( class == P_L0 && part == D_8x16 ) )
  2063. +                            generate_alternate_mvs( h, alt, &(h->mb.trellis.t[col*t_stride+j].block), i_list, i_adj );
  2064. +                        best = 0;
  2065. +                        for ( k = 0; k < h->mb.trellis.breadth; k++ )  {
  2066. +                            for ( m = 0; m < alt_states; m++ ) {
  2067. +                                if ( class == P_8x8 || ( class == P_L0 && part == D_8x16 ) )
  2068. +                                    CP_MV( &(h->mb.trellis.t[col*t_stride+j].block.mv[i_adj]), &(alt[m]) );
  2069. +                                h->mb.trellis.t[col*t_stride+j].costCC = cost_mb( h, row, col, 3, &(h->mb.trellis.t[col*t_stride+j].block), &(h->mb.trellis.t[(col-1)*t_stride+k].block), ( ( row == 0 ) ? NULL : &(h->mb.trellis.mb[offTC+T_P_BEST]) ), ( ( row == 0 ) ? NULL : ( ( col == h->mb.i_mb_width - 1 ) ? &(h->mb.trellis.mb[offTL+T_P_BEST]) : &(h->mb.trellis.mb[offTR+T_P_BEST]) ) ) );
  2070. +                                h->mb.trellis.t[col*t_stride+j].costDC = ( row == h->mb.i_mb_height - 1 ) ? 0 : cost_mb( h, row+1, col, 3, &(h->mb.trellis.mb[offDC+T_P_BEST]), &(h->mb.trellis.mb[offDL+T_P_BEST]), &(h->mb.trellis.t[col*t_stride+j].block), ( ( col == h->mb.i_mb_width - 1 ) ? &(h->mb.trellis.t[(col-1)*t_stride+k].block) : &(h->mb.trellis.mb[offCR+T_P_BEST]) ) );
  2071. +                                h->mb.trellis.t[col*t_stride+j].costDL = ( row == h->mb.i_mb_height - 1 ) ? 0 : cost_mb( h, row+1, col-1, 3, &(h->mb.trellis.mb[offDL+T_P_BEST]), ( ( col == 1 ) ? NULL : &(h->mb.trellis.mb[offDL-mb_stride_col+T_P_BEST]) ), &(h->mb.trellis.t[(col-1)*t_stride+k].block), &(h->mb.trellis.t[col*t_stride+j].block) );
  2072. +                                h->mb.trellis.t[col*t_stride+j].costDR = ( row == h->mb.i_mb_height - 1 ) ? 0 : ( col == h->mb.i_mb_width - 2 ) ? cost_mb( h, row+1, col+1, 3, &(h->mb.trellis.mb[offDR+T_P_BEST]), &(h->mb.trellis.mb[offDC+T_P_BEST]), &(h->mb.trellis.mb[offCR+T_P_BEST]), &(h->mb.trellis.t[col*t_stride+k].block) ) : 0;
  2073. +
  2074. +                                cost = h->mb.trellis.t[(col-1)*t_stride+k].accCost - h->mb.trellis.t[(col-1)*t_stride+k].costDC - h->mb.trellis.t[(col-1)*t_stride+k].costDR + h->mb.trellis.t[col*t_stride+j].costCC + h->mb.trellis.t[col*t_stride+j].costDC + h->mb.trellis.t[col*t_stride+j].costDL + h->mb.trellis.t[col*t_stride+j].costDR;
  2075. +                                if ( cost < min_cost ) {
  2076. +                                    min_cost = cost;
  2077. +                                    best = m;
  2078. +                                    h->mb.trellis.t[col*t_stride+j].parent = k;
  2079. +                                }
  2080. +                            }
  2081. +                        }
  2082. +                        h->mb.trellis.t[col*t_stride+j].accCost = min_cost;
  2083. +                        if ( class == P_8x8 || ( class == P_L0 && part == D_8x16 ) )
  2084. +                            CP_MV( &(h->mb.trellis.t[col*t_stride+j].block.mv[i_adj]), &(alt[T_P_BEST]) );
  2085. +                    }
  2086. +
  2087. +                    offCC += mb_stride_col;
  2088. +                    offCL += mb_stride_col;
  2089. +                    offCR += mb_stride_col;
  2090. +                    offTC += mb_stride_col;
  2091. +                    offDC += mb_stride_col;
  2092. +                    offTL += mb_stride_col;
  2093. +                    offDL += mb_stride_col;
  2094. +                    offTR += mb_stride_col;
  2095. +                    offDR += mb_stride_col;
  2096. +                }
  2097. +
  2098. +                min_cost = INT_MAX;
  2099. +                k = 0;
  2100. +                for ( j = 0; j < h->mb.trellis.breadth; j++ ) {
  2101. +                    if ( h->mb.trellis.t[(h->mb.trellis.length-1)*t_stride+j].accCost < min_cost ) {
  2102. +                        min_cost = h->mb.trellis.t[(h->mb.trellis.length-1)*t_stride+j].accCost;
  2103. +                        k = j;
  2104. +                    }
  2105. +                }
  2106. +
  2107. +                offCC = (row+1) * mb_stride_row - mb_stride_col;
  2108. +                for ( col = h->mb.trellis.length-1; col >= 0; col-- ) {
  2109. +                    t_type = TRELLIS_TYPE( h->mb.trellis.t[col*t_stride+k].block );
  2110. +                    CP_MB( &(h->mb.trellis.mb[offCC+t_type]), &(h->mb.trellis.t[col*t_stride+k].block) );
  2111. +                    CP_MB( &(h->mb.trellis.mb[offCC+T_P_BEST]), &(h->mb.trellis.t[col*t_stride+k].block) );
  2112. +                    k = h->mb.trellis.t[col*t_stride+k].parent;
  2113. +                    offCC -= mb_stride_col;
  2114. +                }
  2115. +
  2116. +                frameCostMVNew = mv_cost( h );
  2117. +                frameCostSADNew = sad_cost( h );
  2118. +
  2119. +                if ( ( frameCostMVNew + frameCostSADNew ) > ( frameCostMVOld + frameCostSADOld ) ) {
  2120. +                    offCC = row * mb_stride_row;
  2121. +                    for ( col = 0; col < h->mb.i_mb_width; col++ ) {
  2122. +                        for ( j = 0; j <= T_P_BEST; j++ )
  2123. +                            CP_MB( &(h->mb.trellis.mb[offCC+j]), &(h->mb.trellis.backup[col*mb_stride_col+j]) );
  2124. +                        offCC += mb_stride_col;
  2125. +                    }
  2126. +                }
  2127. +
  2128. +            }
  2129. +        }
  2130. +    }
  2131. +
  2132. +}
  2133. +
  2134. +void scale_trellis_mv( x264_t *h ) {
  2135. +    mb_t *block;
  2136. +    for ( int row = 0; row < h->mb.i_mb_height; row++ ) {
  2137. +        for ( int col = 0; col < h->mb.i_mb_width; col++ ) {
  2138. +            block = &(h->mb.trellis.mb[row*h->mb.trellis.mb_stride_row+col*h->mb.trellis.mb_stride_col+T_P_BEST]);
  2139. +            if ( ! IS_INTRA( block->class ) ) {
  2140. +                block->mv[0][0][0] <<= 2;
  2141. +                block->mv[0][0][1] <<= 2;
  2142. +                if ( block->part[0] != D_16x16 ) {
  2143. +                    block->mv[0][1][0] <<= 2;
  2144. +                    block->mv[0][1][1] <<= 2;
  2145. +                    if ( block->part[0] == D_8x8 ) {
  2146. +                        block->mv[0][2][0] <<= 2;
  2147. +                        block->mv[0][2][1] <<= 2;
  2148. +                        block->mv[0][3][0] <<= 2;
  2149. +                        block->mv[0][3][1] <<= 2;
  2150. +                    }
  2151. +                }
  2152. +            }
  2153. +        }
  2154. +    }
  2155. +}
  2156. +
  2157. +void write_back_trellis_mv( x264_t *h, int i_mb ) {
  2158. +    mb_t *block = &(h->mb.trellis.mb[i_mb*h->mb.trellis.mb_stride_col+T_P_BEST]);
  2159. +    h->mb.b_skip_mc = 0;
  2160. +
  2161. +    if ( IS_INTRA( block->class ) ) {
  2162. +        h->mb.i_type = block->class;    
  2163. +    }
  2164. +    else {
  2165. +        h->mb.i_type = block->class;
  2166. +        h->mb.i_partition = block->part[0];
  2167. +        
  2168. +        if ( block->class == P_8x8 ) {
  2169. +            h->mb.i_sub_partition[0] = D_L0_8x8;
  2170. +            h->mb.i_sub_partition[1] = D_L0_8x8;
  2171. +            h->mb.i_sub_partition[2] = D_L0_8x8;
  2172. +            h->mb.i_sub_partition[3] = D_L0_8x8;
  2173. +            
  2174. +            x264_macroblock_cache_ref( h, 0, 0, 2, 2, 0, block->ref[0][0] );
  2175. +            x264_macroblock_cache_ref( h, 2, 0, 2, 2, 0, block->ref[1][0] );
  2176. +            x264_macroblock_cache_ref( h, 0, 2, 2, 2, 0, block->ref[2][0] );
  2177. +            x264_macroblock_cache_ref( h, 2, 2, 2, 2, 0, block->ref[3][0] );
  2178. +            x264_macroblock_cache_mv_ptr( h, 0, 0, 2, 2, 0, block->mv[0][0] );
  2179. +            x264_macroblock_cache_mv_ptr( h, 2, 0, 2, 2, 0, block->mv[0][1] );
  2180. +            x264_macroblock_cache_mv_ptr( h, 0, 2, 2, 2, 0, block->mv[0][2] );
  2181. +            x264_macroblock_cache_mv_ptr( h, 2, 2, 2, 2, 0, block->mv[0][3] );
  2182. +        }
  2183. +        else {
  2184. +                
  2185. +            switch ( block->part[0] ) {
  2186. +                
  2187. +                case D_16x16 :
  2188. +                
  2189. +                    x264_macroblock_cache_ref( h, 0, 0, 4, 4, 0, block->ref[0][0] );
  2190. +                    x264_macroblock_cache_mv_ptr( h, 0, 0, 4, 4, 0, block->mv[0][0] );
  2191. +                    break;
  2192. +                    
  2193. +                case D_16x8 :
  2194. +                
  2195. +                    x264_macroblock_cache_ref( h, 0, 0, 4, 2, 0, block->ref[0][0] );
  2196. +                    x264_macroblock_cache_ref( h, 0, 2, 4, 2, 0, block->ref[1][0] );
  2197. +                    x264_macroblock_cache_mv_ptr( h, 0, 0, 4, 2, 0, block->mv[0][0] );
  2198. +                    x264_macroblock_cache_mv_ptr( h, 0, 2, 4, 2, 0, block->mv[0][1] );
  2199. +                    break;
  2200. +                    
  2201. +                case D_8x16 :
  2202. +                
  2203. +                    x264_macroblock_cache_ref( h, 0, 0, 2, 4, 0, block->ref[0][0] );
  2204. +                    x264_macroblock_cache_ref( h, 2, 0, 2, 4, 0, block->ref[1][0] );
  2205. +                    x264_macroblock_cache_mv_ptr( h, 0, 0, 2, 4, 0, block->mv[0][0] );
  2206. +                    x264_macroblock_cache_mv_ptr( h, 2, 0, 2, 4, 0, block->mv[0][1] );
  2207. +                    break;
  2208. +                
  2209. +            }
  2210. +            
  2211. +        }
  2212. +        
  2213. +    }
  2214. +    
  2215. +}
  2216. diff --git a/extras/avisynth_c.h b/extras/avisynth_c.h
  2217. old mode 100644
  2218. new mode 100755
  2219. diff --git a/extras/getopt.c b/extras/getopt.c
  2220. old mode 100644
  2221. new mode 100755
  2222. diff --git a/extras/getopt.h b/extras/getopt.h
  2223. old mode 100644
  2224. new mode 100755
  2225. diff --git a/extras/inttypes.h b/extras/inttypes.h
  2226. old mode 100644
  2227. new mode 100755
  2228. diff --git a/extras/stdint.h b/extras/stdint.h
  2229. old mode 100644
  2230. new mode 100755
  2231. diff --git a/filters/filters.c b/filters/filters.c
  2232. old mode 100644
  2233. new mode 100755
  2234. diff --git a/filters/filters.h b/filters/filters.h
  2235. old mode 100644
  2236. new mode 100755
  2237. diff --git a/filters/video/cache.c b/filters/video/cache.c
  2238. old mode 100644
  2239. new mode 100755
  2240. diff --git a/filters/video/crop.c b/filters/video/crop.c
  2241. old mode 100644
  2242. new mode 100755
  2243. diff --git a/filters/video/depth.c b/filters/video/depth.c
  2244. old mode 100644
  2245. new mode 100755
  2246. diff --git a/filters/video/fix_vfr_pts.c b/filters/video/fix_vfr_pts.c
  2247. old mode 100644
  2248. new mode 100755
  2249. diff --git a/filters/video/internal.c b/filters/video/internal.c
  2250. old mode 100644
  2251. new mode 100755
  2252. diff --git a/filters/video/internal.h b/filters/video/internal.h
  2253. old mode 100644
  2254. new mode 100755
  2255. diff --git a/filters/video/resize.c b/filters/video/resize.c
  2256. old mode 100644
  2257. new mode 100755
  2258. diff --git a/filters/video/select_every.c b/filters/video/select_every.c
  2259. old mode 100644
  2260. new mode 100755
  2261. diff --git a/filters/video/source.c b/filters/video/source.c
  2262. old mode 100644
  2263. new mode 100755
  2264. diff --git a/filters/video/video.c b/filters/video/video.c
  2265. old mode 100644
  2266. new mode 100755
  2267. diff --git a/filters/video/video.h b/filters/video/video.h
  2268. old mode 100644
  2269. new mode 100755
  2270. diff --git a/input/avs.c b/input/avs.c
  2271. old mode 100644
  2272. new mode 100755
  2273. diff --git a/input/ffms.c b/input/ffms.c
  2274. old mode 100644
  2275. new mode 100755
  2276. diff --git a/input/input.c b/input/input.c
  2277. old mode 100644
  2278. new mode 100755
  2279. diff --git a/input/input.h b/input/input.h
  2280. old mode 100644
  2281. new mode 100755
  2282. diff --git a/input/lavf.c b/input/lavf.c
  2283. old mode 100644
  2284. new mode 100755
  2285. diff --git a/input/raw.c b/input/raw.c
  2286. old mode 100644
  2287. new mode 100755
  2288. diff --git a/input/thread.c b/input/thread.c
  2289. old mode 100644
  2290. new mode 100755
  2291. diff --git a/input/timecode.c b/input/timecode.c
  2292. old mode 100644
  2293. new mode 100755
  2294. diff --git a/input/y4m.c b/input/y4m.c
  2295. old mode 100644
  2296. new mode 100755
  2297. diff --git a/output/flv.c b/output/flv.c
  2298. old mode 100644
  2299. new mode 100755
  2300. diff --git a/output/flv_bytestream.c b/output/flv_bytestream.c
  2301. old mode 100644
  2302. new mode 100755
  2303. diff --git a/output/flv_bytestream.h b/output/flv_bytestream.h
  2304. old mode 100644
  2305. new mode 100755
  2306. diff --git a/output/matroska.c b/output/matroska.c
  2307. old mode 100644
  2308. new mode 100755
  2309. diff --git a/output/matroska_ebml.c b/output/matroska_ebml.c
  2310. old mode 100644
  2311. new mode 100755
  2312. diff --git a/output/matroska_ebml.h b/output/matroska_ebml.h
  2313. old mode 100644
  2314. new mode 100755
  2315. diff --git a/output/mp4.c b/output/mp4.c
  2316. old mode 100644
  2317. new mode 100755
  2318. diff --git a/output/output.h b/output/output.h
  2319. old mode 100644
  2320. new mode 100755
  2321. diff --git a/output/raw.c b/output/raw.c
  2322. old mode 100644
  2323. new mode 100755
  2324. diff --git a/tools/checkasm-a.asm b/tools/checkasm-a.asm
  2325. old mode 100644
  2326. new mode 100755
  2327. diff --git a/tools/checkasm.c b/tools/checkasm.c
  2328. old mode 100644
  2329. new mode 100755
  2330. diff --git a/tools/digress/__init__.py b/tools/digress/__init__.py
  2331. old mode 100644
  2332. new mode 100755
  2333. diff --git a/tools/digress/cli.py b/tools/digress/cli.py
  2334. old mode 100644
  2335. new mode 100755
  2336. diff --git a/tools/digress/comparers.py b/tools/digress/comparers.py
  2337. old mode 100644
  2338. new mode 100755
  2339. diff --git a/tools/digress/constants.py b/tools/digress/constants.py
  2340. old mode 100644
  2341. new mode 100755
  2342. diff --git a/tools/digress/errors.py b/tools/digress/errors.py
  2343. old mode 100644
  2344. new mode 100755
  2345. diff --git a/tools/digress/scm/__init__.py b/tools/digress/scm/__init__.py
  2346. old mode 100644
  2347. new mode 100755
  2348. diff --git a/tools/digress/scm/dummy.py b/tools/digress/scm/dummy.py
  2349. old mode 100644
  2350. new mode 100755
  2351. diff --git a/tools/digress/scm/git.py b/tools/digress/scm/git.py
  2352. old mode 100644
  2353. new mode 100755
  2354. diff --git a/tools/digress/testing.py b/tools/digress/testing.py
  2355. old mode 100644
  2356. new mode 100755
  2357. diff --git a/tools/q_matrix_jvt.cfg b/tools/q_matrix_jvt.cfg
  2358. old mode 100644
  2359. new mode 100755
  2360. diff --git a/tools/xyuv.c b/tools/xyuv.c
  2361. old mode 100644
  2362. new mode 100755
  2363. diff --git a/x264.c b/x264.c
  2364. old mode 100644
  2365. new mode 100755
  2366. index 3864bf7..b6a0def
  2367. --- a/x264.c
  2368. +++ b/x264.c
  2369. @@ -966,6 +966,7 @@ static struct option long_options[] =
  2370.      { "input-csp",   required_argument, NULL, OPT_INPUT_CSP },
  2371.      { "input-depth", required_argument, NULL, OPT_INPUT_DEPTH },
  2372.      { "dts-compress",      no_argument, NULL, OPT_DTS_COMPRESSION },
  2373. +    { "me-trellis",  required_argument, NULL, 0 },
  2374.      {0, 0, 0, 0}
  2375.  };
  2376.  
  2377. @@ -1675,7 +1676,7 @@ static int encode( x264_param_t *param, cli_opt_t *opt )
  2378.      FAIL_IF_ERROR2( !h, "x264_encoder_open failed\n" );
  2379.  
  2380.      x264_encoder_parameters( h, param );
  2381. -
  2382. +    
  2383.      FAIL_IF_ERROR2( output.set_param( opt->hout, param ), "can't set outfile param\n" );
  2384.  
  2385.      i_start = x264_mdate();
  2386. diff --git a/x264.h b/x264.h
  2387. old mode 100644
  2388. new mode 100755
  2389. index 9d663f1..37fcd8f
  2390. --- a/x264.h
  2391. +++ b/x264.h
  2392. @@ -162,6 +162,10 @@ typedef struct
  2393.  #define X264_B_PYRAMID_NORMAL        2
  2394.  #define X264_KEYINT_MIN_AUTO         0
  2395.  #define X264_KEYINT_MAX_INFINITE     (1<<30)
  2396. +#define X264_ME_TRELLIS_NONE         0
  2397. +#define X264_ME_TRELLIS_DIA          1
  2398. +#define X264_ME_TRELLIS_ESA          2
  2399. +#define X264_ME_TRELLIS_COMBINED_DIA 3
  2400.  
  2401.  static const char * const x264_direct_pred_names[] = { "none", "spatial", "temporal", "auto", 0 };
  2402.  static const char * const x264_motion_est_names[] = { "dia", "hex", "umh", "esa", "tesa", 0 };
  2403. @@ -332,6 +336,8 @@ typedef struct x264_param_t
  2404.          float        f_psy_rd; /* Psy RD strength */
  2405.          float        f_psy_trellis; /* Psy trellis strength */
  2406.          int          b_psy; /* Toggle all psy optimizations */
  2407. +        int          i_me_trellis; /* Type of trellis to use for me */
  2408. +        int          b_cache_sads; /* Cache SAD scores during motion estimation */
  2409.  
  2410.          /* the deadzone size that will be used in luma quantization */
  2411.          int          i_luma_deadzone[2]; /* {inter, intra} */
  2412. diff --git a/x264cli.h b/x264cli.h
  2413. old mode 100644
  2414. new mode 100755
  2415. diff --git a/x264dll.c b/x264dll.c
  2416. old mode 100644
  2417. new mode 100755
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!